Studies of pyrolysis and oxidation of methyl formate using molecular beam mass spectrometry

Paper 070RK-0168 0168 Topic: Reaction Kinetics 8 th U. S. National Combustion Meeting Organized by the Western States Section of the Combustion Institute and hosted by the University of Utah May 19-22, 2013 Studies of pyrolysis and oxidation of methyl formate using molecular beam mass spectrometry Naoki Kurimoto, Xueliang Yang and Yiguang Ju Department of Mechanical and Aerospace Engineering, Princeton University, Princeton NJ-08542, United States Molecular beam mass spectrometry with electron impact ionization coupled with a quartz flow reactor has been employed to study the pyrolysis and the oxidation mechanism of methyl formate (MF) at atmospheric pressure. The measurement was carried out with a mixture of gas phase at 5000 ppm MF in Helium and Argon dilution in the temperature range from 500 K to 1000 K with the reaction residence time of 600 milliseconds. Important stable species such as methanol, carbon monoxide, carbon dioxide, formaldehyde, water as well as radical species such as HCO were quantified in the mass spectrum. Effect of electron impact fragmentation of MF and methanol on species measurements is calibrated and subtracted. Experimental uncertainty is estimated to be approximately 10% for MF at 95% confidence. The experimental results showed that pyrolysis and oxidation take place for the temperature higher than 700 K with the increase of products such as methanol, carbon monoxide and carbon dioxide. Numerical simulations using a Princeton Ester-Mech kinetic model and a LLNL model have been performed. The results show that the model under-predicts the formation of CO for the pyrolysis study and CO 2 for the oxidation study, respectively. The discrepancy between the measured and the predicted profile implies that the reaction pathways for hydrogen abstractions from MF forming CH 2 OCHO and CH 3 OCO are under-predicted in the modeling. 1. Introduction Fuel diversity is essential to a sustainable energy society. Fatty acid methyl esters made from plant oil such as palm oil, soybean oil, rapeseed oil, sunflower oil and Jatropha oil are the major compositions of biodiesels. Production of biodiesels is under way in earnest since the finite nature and the price increase of the conventional fossil fuel is now widely recognized. Nevertheless, to be a major transportation fuel, biodiesels still face several technical challenges. One of the challenge is that biodiesel is much more expensive than conventional fossil fuels. Another challenge is that they require conventional engines to be extensively redesigned to meet emission regulations. In order to overcome these disadvantages and make biodiesels highly competitive in the market, it is necessary to understand their reaction kinetics in detail so that we could actively utilize its chemical properties to improve an engine performance with a resultant economic advantage. Especially, development of a validated methyl ester kinetic model is practically important to facilitate parametric studies with conventional engines (Herbinet et al., 2008; Diévart et al., 2012 and 2013; Dooley et al., 2008). Methyl formate (MF: CH 3 OCHO) is the simplest methyl ester and thus is an ideal molecule to isolate the effect of the ester functionality on its combustion process. Kinetic modeling or experimental studies of the methyl formate pyrolysis and oxidation have been carried out in the past decade, and the kinetic models have reproduced experimental results (Fisher et al., 2000). Especially, Dooley et al. (2010) developed the kinetic model and could reproduce the production of stable intermediate species formed in a flow reactor at a fixed temperature. However, in their experiment, significant amount of major intermediate species were detected downstream a diffuser of the flow reactor due to catalytic reaction, and thus the initial mixture was not well characterized. In addition, since their experiment was carried out at the fixed temperature of 975 K, experimental data across a range of temperature lower than 1000 K is not yet adequate to

validate those kinetic models. Later, Diévart et al.(2013) performed the flame extinction limit studies of one sets of small methyl esters, and concluded that methyl formate combustion character heavily relies on the reaction kinetics of methanol which is one of the main consumption pathway for methyl formate. Therefore, a more careful combustion kinetics model of methanol and methyl formate including theoretical assessment on the relevant reaction rates and more serious test against the currently available experimental observation has been developed (Diévart et al., 070RK-0276). The final goal of the present study is to provide reliable experimental data of the major intermediate species such as methanol, carbon monoxide, carbon dioxide, methane and formaldehyde, and provide constraints on the prediction of branching reaction pathways of methyl formate. 2. Experimental methods 2.1 Flow facility and Flow reactor A reactant mixture supplied to an atmospheric flow reactor consists of helium (He), argon (Ar), oxygen (O 2 ) and methyl formate (MF). The helium, the argon and the oxygen were supplied from industrial-grade compressed gas cylinders, and each flow rate was regulated by a mass flow controller (MKS instruments, 1179A Mass-Flo). These gases were mixed to form a career gas. The career gas was preheated up to 306 K in order to promote vaporization of the methyl formate. The methyl formate (Sigma-Aldrich, 99% purity) was injected into the preheated career gas from a fine tube with the inner diameter of 0.3 mm. The small inner diameter allows rapid preheating and reduces timescale of temperature of non-uniformity. The flow rate of the methyl formate was controlled by a syringe pump with a 200 ml stainless steel syringe (Harvard Apparatus, PHD2000 HPSI). This injection part was placed at 700 mm upstream of a flow reactor to ensure homogeneous mixing. The mixture flowed to the flow reactor through a stainless tube at a room temperature, and thus any reaction upstream the flow reactor should be negligible. A cylindrical quartz tube of 17 mm inner diameter and 355 mm in length was employed as the flow reactor. The reactor was tightly jacketed within a copper sleeve, and the assembly was placed inside an oven to generate a uniform temperature profile in the reactor. The reactor temperature was controlled by a PID temperature controller with a K-type thermocouple installed inside at the streamwise center of the reactor. In order to heat the reactant mixture up to the set temperature for the flow reactor rapidly, a preheating section with the inner diameter of 2 mm was built at the entrance of the reactor. The residence time of the mixture in the preheating section is less than 1% of the total residence time. In this study, a He/Ar/O 2 /MF mixture of 0.945/0.05/0.0/0.005 and 0.94/0.5/0.05/0.05 mole fraction was employed for the pyrolysis experiment and the oxidation experiment, respectively. Experiments were carried out at atmospheric pressure for the temperature ranging from 500 K to 1000 K with the variation by less than +-3 K. The reaction residence time was set to be 600 milliseconds and the corresponding flow rate was adjusted in order to keep the constant residence time when the set reactor temperature was varied. 2.2 Concentration measurements through EI-MBMS Electron-Ionization Molecular Beam Mass Spectrometry (EI-MBMS) was employed to quantify the concentration of reactants, intermediate species, and products. Figure 1 shows the schematic of the molecular beam mass spectrometry. The EI-MBMS consists of a sampling-chamber with a quartz sampling nozzle and a skimmer, an ionization-chamber, an electron gun (e-gun) and a time-of-flight mass spectrometer. Detailed descriptions of the instrument are given elsewhere (Guo et al., 2013). The pressures in the sampling-chamber, the ionization-chamber and the time-of-flight mass spectrometer reached 3 10-4 Torr, 5 10-6 Torr and 5 10-7 Torr, respectively. In order to suppress a fragmentation of target species, ionization electron energy of the e-gun was set to be as small as 12 ev with the emission current of 0.1 ma. Moreover, voltage parameters of the time-of-flight mass spectrometer, such as a pulsed ion-extraction field, were optimized in such a way that each detected peak in a spectrum became a Gaussian profile that provides the minimum mass resolution among all parameter combinations examined. The nominal mass resolutions, m/ m, based on full width at half maximum were thus 490, 800, 710, and 830 for He, Ar, O 2 and MF peaks, respectively. The maximum background random noise was as small as 2 counts, and the resulting signal-to-noise ratio was ensured larger than 20 for all detected peaks. In the present study, since the mass resolution was approximately 710 at the mass-to-charge ratio (m/z) of 32, a methanol peak at m/z = 32.026, which is one of the most important species in MF pyrolysis and oxidation, somewhat 2

overlapped an oxygen peak at m/z = 31.989. In order to separate the methanol and the oxygen peaks, two Gaussian functions were employed to fit the spectrum around m/z = 32. The signal of each species was defined as the integral from positive infinity to negative infinity of the Gaussian function. The uncertainty due to the fitting itself was statistically calculated to be 6%. The concentration calibration of the EI-MBMS was conducted by flowing mixtures with known compositions. Detailed descriptions of the calibration process are given elsewhere (Guo et al., 2013). In the present study, the calibrations were directly carried out for MF, O 2, CH 3 OH, CH 4, CO and CO 2, while the calibration using the experimentally determined mass discrimination factors and the ionization cross sections obtained from the database of National Institute of Standard and Technology (NIST) were carried out for CH 2 O, HCO and H 2 O. For the reference gas, Ar 5.0% was employed, and all ion signals were normalized by the intensity of argon ion signal in order to correct the signal drop due to the decrease of the amount of the sampling gas when the temperature is increased. EI fragmentations from MF and CH 3 OH were not negligible in the spectrum analysis. That is, MF fragmented into CH 3 OH, CH 3 O, CH 2 O, HCO, CO and CH 3 by the excessive energy due to an electron impact, while CH 3 OH fragmented into CH 3 O, CH 2 O, HCO and CH 3. Thus, in the present study, the effect of the fragment ions from the MF and the CH 3 OH ions on the target species was eliminated following to the equation, (1) where X, C and S denote a concentration vector, a calibration matrix, and a signal vector, respectively. Each column of the matrix C is the fragment spectrum experimentally obtained in the calibration. Measurement uncertainty was systematically analyzed following to the guideline provided by NIST (Taylor and Kuyatt, 1994). The uncertainty budget showed that the random effect of the signal intensity normalized by Ar signal is a dominant source in the total uncertainty. Thus, in the present study, a total of three experiments were carried out to calculate the average, and then the total uncertainty for MF was estimated to be 10% at 95% confidence. 2.3 Simulation Two kinetic models for MF, the Princeton Ester-Mech (Diévart et al. 2013) and the LLNL model (Westbrook et al. 2009) were employed. Model predictions were obtained by using the software package Cantera (Goodwin, 2001-2005). A reactor model was solved under a constant temperature condition for the reaction residence time of 600 milliseconds, assuming that the temperature profile is uniform along the axis of the flow reactor. 3. Results and Discussion 3.1 Mass spectrum Figure 2 shows a typical mass spectrum for He/Ar/O 2 /MF (0.94/0.05/0.005/0.005) mixture at T=1000 K with the residence time of 600 milliseconds. Distinct mass peaks were detected at the mass charges of 60 (MF: CH 3 OCHO), 44 3

(CO 2 ), 40 (Ar), 32 (CH 3 OH and O 2 ), 31 (CH 3 O), 30 (CH 2 O), 29 (HCO), 28 (CO), 18 (H 2 O), 15 (CH 3 ), 16 (CH 4 ), 4 (He). It should be noted that we did not find a distinct peak at m/z = 59 (CH 2 OCHO and CH 3 OCO), which are produced by the hydrogen abstraction from MF. This indicates that the decomposition rate of the CH 2 OCHO and the CH 3 OCO radical into CH 2 O + CHO, CH 3 O + CO and CH 3 + CO 2 is fast enough, and the concentration of these species were too low to be detected at the sampling point. It should be noted again that each detected peak was not only from a combustion product but also from a fragment of larger species than the target species. Thus, hereafter, the contributions of the fragment ions from the MF, the CH 3 OH, the CO 2 and the O 2 ions were eliminated from each peak following to Eq. (1). 3.2 Pyrolysis 3.2.1 Measured and predicted species profiles Figure 3 shows the profiles of the measured and the predicted major species for the pyrolysis conditions of 0.945/0.05/0.005 He/Ar/MF at atmospheric pressure with residence time of 600 milliseconds. As shown in Fig. 3 (a) and (b), although the concentration of MF was almost constant at the temperature lower than 700 K, a very small amount of CH 3 OH and CO formation was detected at 700 K. This is probably due to the surface catalytic reaction observed at the inlet of a flow reactor in Dooley et al., (2010). The concentration of MF started to decrease gradually with the increase of the major intermediate species at the temperature as low as 700 K. The concentration decreased rapidly at T = 900 K - 925 K with the significant increase of CO concentration, and subsequently decreased gradually again at T = 925K - 1000 K. Both models by Westbrook et al. (2009) and Diévart et al. (2013) show the similar trend. However, both models predict that the MF concentration starts to decrease at higher temperature and drops more rapidly with the temperature increase. This indicates that these models under-predict the reactivity of the fuel consumption at the low temperature range while over-predict the reactivity at the high temperature range. The measured concentrations of CH 3 OH and CO started to increase at the temperature as low as 700 K. It should be noted that the measured CO concentration was approximately at the same level as the CH 3 OH concentration at T = 800 K - 850 K, while the CO concentration became more than twice of the CH 3 OH concentration at the temperature higher than 900 K where the CH 3 OH concentration still increases. As described later, this trend implies that, before CH 3 OH starts to decompose, the hydrogen abstraction from MF forming the CH 2 OCHO radical becomes more dominant than the concerted elimination reaction of MF = CH 3 OH + CO. On the other hand, the model prediction shows that the CH 3 OH and the CO concentration start to increase at 850 K. This means that the models under-predict the reactivity of the MF decomposition as described above. In addition, the model of Westbrook et al. (2009) shows a peak of the CH 3 OH concentration at T = 973 K, where both of the experiment and the model prediction of Diévart et al. (2013) show that the concentration still increases. This indicates that the model of Westbrook et al. (2009) over-predicts of the decomposition rate of CH 3 OH. Moreover, the model prediction of 4

Diévart et al. (2013) shows that the CO concentration is at similar level to the CH 3 OH concentration until the CH 3 OH concentration has a peak at T = 1023 K. The model prediction of Westbrook et al. (2009) reproduces the experimental trend better since it shows that CO concentration is larger than the CH 3 OH concentration before the CH 3 OH concentration has a peak at T = 973 K. The measured CO 2 concentration was as low as 40 ppm - 100 ppm while the CH 4 concentration was 30 ppm 50 ppm over the temperature range examined (Fig. 3c). On the other hand, both models over-predict both of the CO 2 and CH 4 concentrations for the temperature higher than 925 K. In the present measurement, the concentration increase of CH 2 O was very slightly observed at the temperature higher than 975 K while that of HCO radical was distinctly observed from the temperature as low as 800 K (Fig. 3d). The reason that the measured CH 2 O concentration was much smaller than that of the HCO radical is probably because most of the CH 2 O ions decomposed into its fragment ions of HCO radical due to the excessive energy of an electron impact. In the concentration profile, the HCO radical concentration, which also represents the CH 2 O concentration, increased from T = 800 K to 850 K and subsequently became almost constant. This indicates that the HCO radical is 5

kept accumulated before the production rate of the HCO radical became in balance with its decomposition rate into CO and H. On the other hand, it is found that both model predictions of Westbrook et al. (2009) and Diévart et al. (2013) over-predict the CH2O concentration at its peak approximately by a factor of two and four, respectively. Since the CH2OCHO radical, which decomposes into CO through CH2O and HCO, should not be over-predicted in the modeling as describe above, the reason that the models over-predict the peak concentration of CH2O is probably because they under-predict the CH2O consumption. 3.2.2 Reaction path analysis Figure 4 shows the diagram of the reaction rate of progress calculated with the model of Diévart et al. (2013). The temperature is set at 973 K, since a significant discrepancy between the experimental results and the model predictions was observed in the CH3OH and the CO formation. The model shows that MF is consumed almost exclusively by concerted elimination reaction forming CH3OH + CO (88.6%), CH4 + CO2 (7.6%) and CH2O + CH2O (3.4%). Contributions of hydrogen abstraction from MF forming CH2OCHO and CH3OCO, which produce CO and CO2 through the decomposition, are as small as 0.3% and 0.1%, respectively. In the experiment, the CO concentration (2527 ppm) was more than twice of the CH3OH concentration (1052 ppm) at T = 975 K. The fact the measured CO concentration was higher than its counterpart species of the concerted elimination reaction implies that the production of the CH2OCHO radical is underestimated in the model. This conjecture is supported by the measured HCO and CH2O concentration, which is about four times larger than the model prediction at T = 973 K. Therefore, it is concluded that the reaction pathway for the hydrogen abstraction from MF on the methyl site is significantly underestimated in the modeling. 3.3 Oxidation 3.3.1 Measured and predicted species profiles Figure 5 shows the profiles of the measured and the predicted major species for the oxidation conditions of 0.940/0.05/0.005/0.005 He/Ar/O2/MF at atmospheric pressure with residence time of 600 milliseconds. As shown in Fig. 5 (a) and (b), the concentration of MF were almost constant with a little increase of CH3OH and CO for the temperature 6

lower than 700 K, and started to decrease with the increase of the CH 3 OH and CO concentrations at the temperature higher than 700 K. This result is in the same trend as the pyrolysis study, indicating that there is no low temperature chemistry for MF. However, in the MF oxidation process, the MF concentration decreased more rapidly when the temperature was increased to 925 K - 1000 K. This result is reasonable because the oxygen addition increases hydrogen abstraction reaction from MF, and the resulting abundant radicals such as OH and HO 2 should further promote the MF consumption. The model predictions show a similar trend, but once again they over-predict the temperature where the MF consumption starts and also over-predict the sensitivity of the MF consumption to the temperature increase. As shown in Figure 5b, the measured CH 3 OH concentration started to increase at the temperature as low as 700 K just like the pyrolysis study. However, unlike pyrolysis, the CH 3 OH concentration had a local peak around T = 925 K, and subsequently decreased at higher temperature. This indicates that CH 3 OH oxidation starts at this temperature. Moreover, the CO concentration was at a similar level to the CH 3 OH concentration until the CH 3 OH concentration had a peak around T = 925 K. Both models of Westbrook et al. (2009) and Diévart et al. (2013) reproduce the experimental trend that the CH 3 OH concentration has a local peak, although they over-predict the peak location and slightly under-predict the peak value. It should also be noted that, unlike the pyrolysis study, the model of Westbrook et al. (2009) over-predict the CO 7

concentration until the CH3OH concentration has a peak at T = 943 K. On the other hand, the model of Diévart et al. (2013) predicts the experimental trend better in CO and CH3OH. The CO concentration is at a similar level to the CH3OH concentration until the CH3OH concentration has a peak at T = 963 K. As shown in Figure 5c, the measured CO2 concentration started to increase at the same temperature as the CO concentration increases. The measured CO2 concentration at T = 925 K is 646 ppm, which is much higher than that of CH4. Two reasons are conjectured for the CO2 increase. One is that it comes from the decomposition of the CH3OCO radical since the increase of the CH4 concentration, which is the counterpart species of the concerted elimination reaction forming CH4 + CO2, starts at the higher temperature than that of CO2. The other is that it comes from the oxidation of CO, which is produced through the decomposition of CH2OCHO radicals. On the other hand, in the model predictions, the CO2 concentration is under-predicted before the CH3OCO starts to be oxidized while over-predicted after that. The CH4 concentration is reproduced well in the model predictions. The measured H2O concentration increased with the temperature increase unlike the pyrolysis study, implying that there were abundant OH radicals. This supports the above-mentioned conjecture that the CO oxidation was promoted in the oxidation. Like the CO2 concentration, the models also under-predicts the H2O concentration before the CH3OCO starts to be oxidized while over-predicted after that. The measured CH2O and the HCO concentration show the same trend as the pyrolysis study except that the CH2O concentration is higher in the oxidation than in pyrolysis (Fig. 5d). On the other hand, the models predict the peak value better in the oxidation study than in the pyrolysis study, implying that they predict the CH2O consumption better. 3.3.2 Reaction path analysis Figure 6 shows the diagram of the reaction rate of progress calculated with the model of Diévart et al. (2013) for the oxidation. The temperature is set at 923 K, at which the predicted CH3OH concentration becomes half of its peak value, since a significant discrepancy between the experimental results and the model predictions was observed in the CO2 concentration. The model shows that MF is consumed by concerted elimination reaction forming CH3OH + CO (88.5%), CH4 + CO2 (7.6%) and CH2O + CH2O (3.4%). Contributions of hydrogen abstraction from MF forming CH2OCHO and 8

CH 3 OCO are as small as 0.3% and 0.2% respectively, which is very similar to the pyrolysis study although OH radical plays a more role in the hydrogen abstraction. Then, the CO concentration is only 1.5 % larger than the CH 3 OH concentration. On the other hand, the measured CO 2 concentration were much higher than the CH 4 concentrations at T = 850 K, at which the measured CH 3 OH concentration becomes half of its peak value. Moreover, the measured CO 2 concentration in the oxidation study is much higher than in the pyrolysis study, while the CO concentration is at similar level to that of the pyrolysis. The fact that the CO 2 concentration was much higher than its counterpart species of the concerted elimination reaction implies that the production of the CH 3 OCO radicals is underestimated in the model. Therefore, it is concluded that the reaction pathway for the hydrogen abstraction from MF on the ester site should be more intensified in the kinetic models for the oxidation study. 4. Conclusions The pyrolysis and the oxidation of methyl formate (MF) have been studied in a quartz flow reactor at atmospheric pressure over the temperature range of 500 K to 1000 K with the aid of molecular beam mass spectrometry. For the pyrolysis study, a discrepancy between the experimental results and the model predictions was observed especially in the CH 3 OH/CO formation before the CH 3 OH starts to decompose: The measured CO concentration is more than twice of the CH 3 OH concentration. The fact that the measured CO concentrations was larger than its counterpart species of the concerted elimination reaction forming CH 3 OH + CO leads us to the conclusion that the reaction pathway for the hydrogen abstraction from MF on the methyl site is underestimated in the modeling. For the oxidation study, a discrepancy between the experimental results and the model predictions was observed especially in the CO 2 formation before the CH 3 OH starts to be oxidized: The measured CO 2 concentration was much higher than the CH 4 concentration. Moreover, the measured CO 2 concentration in the oxidation was also much larger than in the pyrolysis, while the measured CO concentration was at a similar level to that in MF pyrolysis. From the experimental measurements, it can be concluded that, especially for the oxidation study, the reaction rates for the hydrogen abstraction on the ester site should be higher than those in the kinetic models. Acknowledgements This work is supported as part of the Combustion Energy Frontier Research Center, funded by the United States Department of Energy, office of Science, Office of Basic Energy Sciences under Award Number De-SC0001198. This work is also supported by DENSO CORPORATION. References Diévart, P.; Won, S. H.; Dooley, S.; Dryer, F. L.; Ju, Y. A kinetic model for methyl decanoate combustion (2012). Combust. Flame 159 (5), 1793-1805. Diévart, P.; Won, S. H.; Gong, J.; Dooley, S.; Ju, Y. A comparative study of the chemical kinetic characteristics of small methyl esters in diffusion flame extinction (2013). Proc. Combust. Inst. 34 (1), 821-829. Dooley, S.; Curran, H. J.; Simmie, J. M. Autoignition measurements and a validated kinetic model for the biodiesel surrogate, methyl butanoate (2008). Combust. Flame 153 (1 2), 2-32. Fisher, E.M.; Pitz, W.J.; Curran, H.J.; Westbrook, C.K. Detailed chemical kinetic mechanisms for combustion of oxygenated fuels (2000). Proc. Combust Inst., 28, 1579-1586. Dooley, S.; Burke, M.P.; Chaos, M.; Stein, T.; Dryer, F.L.; Zhukov, V.P.; Finch, O.; Simmie, J.M.;Curran, H.J. Methyl formate oxidation: Speciation data, laminar burning velocities, ignition delay times, and a validated chemical kinetic model (2010). Int. J. Chem. Kinet., 42, 527 529. Goodwin, D.G. Cantera, http://www.cantera.org, 2001 2005. Guo, H.; Sun, W.; Haas, F.M.; Farouk, T.; Dryer, F.L.; Ju, Y. Measurements of H2O2 in Low Temperature Dimethyl Ether Oxidation (2013). Proc. Combust Inst, 34, 573-581. 9

Herbinet, O.; Pitz, W. J.; Westbrook, C. K. Detailed chemical kinetic oxidation mechanism for a biodiesel surrogate (2008). Combust. Flame 154 (3), 507-528. Taylor, B.N.; Kuyatt, C.E. Guidelines for evaluating and expressing the uncertainty of NIST measurement results (1994). NIST Technical Note 1297, 1994 Edition. 10