Autoignition Study of 1-Methylnaphthalene in a ... - ACS Publications

Dec 15, 2016 - 1-Methylnaphthalene (1-MN), also known as α-methylnaphthalene, is a substituted diaromatic hydrocarbon that is widely used as one of t...
0 downloads 10 Views 2MB Size
Download PDF
Article pubs.acs.org/EF

Autoignition Study of 1‑Methylnaphthalene in a Rapid Compression Machine Goutham Kukkadapu* and Chih-Jen Sung Department of Mechanical Engineering, University of Connecticut, Storrs, Connecticut 06269, United States ABSTRACT: 1-Methylnaphthalene (1-MN), also known as α-methylnaphthalene, is a substituted diaromatic hydrocarbon that is widely used as one of the representative aromatic hydrocarbons in surrogate fuels of diesel. The ignition characteristics of 1MN have been investigated in the present study using a rapid compression machine (RCM) at compressed pressures of PC = 15− 40 bar, low-to-intermediate compressed temperatures of TC = 837−980 K, and varying equivalence ratios of ϕ = 0.5, 1.0, and 1.5 in air. The current RCM ignition delay measurements were found to complement the existing shock tube ignition delay data in the literature. The performance of two of the latest literature chemical kinetic models in estimating the ignition characteristics of 1-MN has been assessed and discussed. In addition, reaction path analyses and brute force sensitivity analyses have been conducted to identify the important oxidation pathways of 1-MN described by the two literature models. Based on the results from these chemical kinetic analyses, the improvements to the 1-MN model are suggested and discussed. oxidation of 1-MN,8 which was found to reproduce the experimental results to a good degree of agreement. Pfahl et al.9 studied the autoignition of stoichiometric 1-MN + air mixtures behind reflected shock waves at a nominal pressure of 13 bar and temperatures of 840−1300 K. Wang et al.10 studied the autoignition of 1-MN behind reflected shock waves at pressures of 8−45 atm, varying equivalence ratios of ϕ = 0.5−1.5, and temperatures in the range of 1032−1445 K. A chemical kinetic model of 1-MN was also proposed in ref 10 that was found to predict the ignition delays and speciation measurements to good agreement. Bounaceur et al.11 proposed a chemical kinetic model of 1-MN and validated their model against the literature data. Narayanaswamy et al.12 proposed a 1-MN model as well and validated their model against the flow reactor data of Shaddix et al.6 and the ignition delay data from Pfahl et al.9 Review of the literature studies shows that the oxidation of 1MN has not been investigated in detail at low-to-intermediate temperatures (e.g., 600−1000 K) and the performance of the available chemical kinetic models at low-to-intermediate temperatures is not known. Realizing this, the autoignition characteristics of 1-MN have been studied in a rapid compression machine (RCM) herein. The current study intends to provide ignition delay information on 1-MN at low-to-intermediate temperatures with varying pressures and equivalence ratios. In addition, the ignition delays measured in the current study were used to assess the performance of the available models, and such a comparison would help understand the areas where the improvements to the literature models can be made. In the following sections, the experimental facility and the test conditions will be first described. Next, the experimental results will be presented and compared with the literature data. Subsequently, the simulated results using two chemical kinetic

1. INTRODUCTION Since aromatic hydrocarbons constitute about 30% (by weight) of diesel fuels,1,2 understanding their chemical kinetics is necessary for developing comprehensive oxidation models to describe ignition and combustion of surrogate fuels of diesel. 1Methylnaphthalene (1-MN), also known as α-methylnaphthalene, is a substituted diaromatic hydrocarbon and has been used as a surrogate component, representative of aromatic hydrocarbons, for diesel in several studies.2−5 1-MN is popularly used in diesel surrogate formulations, as it was earlier used as a reference fuel for determining the cetane rating of a fuel. Recognizing the importance of understanding the autoignition characteristics of 1-MN, we aim to investigate its ignition delay times at engine-relevant conditions. In the following, the chemical kinetic studies on 1-MN reported in the literature are reviewed, including those conducted in a flow reactor,6,7 a jet stirred reactor,8 and shock tubes.9,10 Shaddix et al.6 studied the oxidation of 1-MN at atmospheric pressure in a flow reactor over initial temperatures of 1170− 1200 K and at varying equivalence ratios of ϕ = 0.6−1.5. Based on the analysis of their measured speciation profiles, Shaddix et al.6 postulated that the oxidation of 1-MN is dominated by the reactions of the 1-naphthylmethyl radical while the reactions of 1-MN with oxygen atom may also be significant. In their subsequent study, Shaddix et al.7 showed through the computed fuel decay path fluxes that, under flow reactor conditions, abstraction of a “benzylic” hydrogen from the methyl side chain dominates consumption of 1-MN, while oxygen atom addition to the aromatic ring, displacement of the side chain by hydrogen atom, and homolytic decay also contribute significantly. In addition, changes in the 1-MN decay mechanism under higher temperature combustion environments and the oxidation mechanisms of even larger polycyclic aromatic hydrocarbons were postulated in ref 7. Mati et al.8 studied the oxidation of diluted 1-MN + O2 + N2 mixtures in a jet stirred reactor for pressures of 1−10 atm, temperatures of 800−1421 K, and equivalence ratios of ϕ = 0.5−1.5. The authors also developed a chemical kinetic model to describe the © XXXX American Chemical Society

Received: July 5, 2016 Revised: December 15, 2016 Published: December 15, 2016 A

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels models taken from the literature will be shown and compared with the current experimental data sets. Based on the comparison, the results of reaction path analyses and brute force sensitivity analyses conducted at selected conditions will be discussed to identify the controlling chemistry, and the potential areas required further studies to improve the comprehensiveness of the literature models.

2. EXPERIMENTAL SPECIFICATIONS 2.1. Experimental facility. A rapid compression machine has been used in the current autoignition study of 1-MN. The test mixture is compressed to elevated pressure and temperature using a fast moving creviced piston driven pneumatically. Toward the end of compression, the piston is brought to rest using a hydraulic pin-groove mechanism. The compression time generally varies between 30 and 45 ms, but the majority of the rise in pressure occurs during the last 6 ms of the compression stroke. This RCM is versatile in its design, offering optical accessibility, an ability to conduct rapid gas sampling measurements, and a wide range of geometric compression ratios. The geometric compression ratio (CR) is varied by changing the stroke length or/and the end of the compression (EOC) clearance length, as defined in the following. CR = 1 +

Stroke length EOC Clearance length

Figure 1. Plot showing the definition of ignition delay used in the current study. of four consecutive runs was found to be less than 15% of the reported ignition delay that corresponds to the experimental run whose ignition delay value is close to the mean of the consecutive runs. Figure 2 shows the typical experimental repeatability of the current ignition delay measurements.

(1)

In the current design, the stroke length can be varied between 177.8 and 254 mm in increments of 6.35 mm while the EOC clearance length can be varied between 14 and 56 mm in increments of 1 mm. Thus, by changing the EOC clearance length and stoke length, the operating geometric compression ratio can be varied from 4 to 20. The change in CR helps in scanning across a wide range of pressure/ temperature conditions at the end of compression, represented by compressed pressure (PC) and compressed temperature (TC). By changing the intake pressure (P0) of the homogeneous mixture of fuel and oxidizer, PC can be varied in the range of 7−70 bar. In addition, the dynamic pressure in the reaction cylinder is measured using a Kistler 6125C transducer in conjunction with a Kistler 5010B charge amplifier. The intake homogeneous mixture of fuel and oxidizer is prepared in a 17.47-L stainless steel tank equipped with heating tapes, magnetic stirrer, and pressure relief disc. The heating tapes are used to heat the mixture to the desired intake temperature (T0) while the magnetic stirrer placed at the bottom of the tank aids in proper mixing. In the current experiments, the mixture is prepared at room temperature by filling the mixing chamber with fuel, followed by the oxidizer gases. The liquid fuel (i.e., 1-MN) is injected directly into the mixing chamber using a glass syringe. The mass of the liquid fuel injected into the chamber is measured using a weigh balance which has a least count of 0.01 g. For the mixture composition considered here, the mass of 1MN injected into the mixing chamber varied in the range of 1.8−3.0 g. The oxidizer gases are filled into the mixing chamber on a barometric measure, after injection of the liquid fuel. The oxidizer gases used in the current study were oxygen and nitrogen in the molar proportion of O2:N2 = 1:3.76. Ultrahigh purity (>99.99%) grade oxygen and nitrogen obtained from Airgas were used in the current experiments. Once the mixture of fuel and oxidizer is prepared, the heaters and the magnetic stirrer are activated. The mixing chamber and the whole RCM facility are heated to the desired preheat temperature for about 4 h before the start of experiments to attain homogeneity. Further details of the current RCM design and procedures can be found in ref 13. 2.2. Definition of RCM ignition delay. The ignition delay definition used in the current study is best illustrated by Figure 1, identified using the local maximum of the instantaneous derivative of the pressure trace. The ignition delays reported here were measured relative to the EOC, which is set as t = 0 ms, as shown in Figure 1. It is noted that the uncertainties in mixture preparation, P0, T0, dynamic pressure trace, etc. could affect the ignition delay measurements. Typically, the scatter in ignition delays determined using a minimum

Figure 2. Plot showing the repeatability of the current experiments. 2.3. Determination of compressed temperature (TC). The temperature of the test mixture at the end of compression was estimated using “adiabatic core” hypothesis. According to this hypothesis, the core of the mixture is compressed adiabatically and the mixture temperature at the end of compression can be deduced from

∫T

Tc

0

⎛P ⎞ γ dT = ln⎜ c ⎟ γ−1 T ⎝ P0 ⎠

(2)

where PC, P0, and T0 are measured values and γ is the ratio of specific heats, which is temperature dependent. Estimation of compressed temperature using this hypothesis needs the thermodynamic information on the test mixture. The thermodynamic properties of 1-MN, O2, and N2 were taken from the model of Narayanaswamy et al.12 The compressed temperatures deduced using the thermodynamic properties from the model of Wang et al.10 were found to be identical to those deduced using the model of Narayanaswamy et al.12 2.4. Test conditions. Table 1 summarizes the test conditions, including equivalence ratio (ϕ), mole percentage of 1-MN (X1‑MN), mole percentage of O2 (XO2), mole percentage of N2 (XN2), and PC, investigated in the current study. The test conditions were designed to study the effects of pressure and equivalence ratio on the autoignition of 1-MN over a range of compressed temperatures for TC = 837−980 K. The pressure and equivalence ratios were also chosen in order to provide ignition delay data at lower temperatures that will complement the existing shock tube data of Pfahl et al.9 and Wang et al.10 All the B

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 4 shows the Arrhenius plots of the ignition delays obtained in this study. As mentioned earlier, the uncertainty in the reported ignition delay is about 15%. The ignition delays shown in Figure 4 are seen to decrease monotonically with increasing temperature. Also shown in Figure 4 are the ignition delay data from the shock tube experiments of Pfahl et al.9 and Wang et al.10 It is noted that the shock tube ignition delay data have been scaled to the current experimental conditions using the pressure exponent (∼P−0.9) suggested by Wang et al.10 As seen from Figure 4, the current ignition delay measurements complement and compare well with the shock tube data in the literature. Figure 4 also elucidates the importance of the current RCM study in providing the ignition delay measurements at low-tointermediate temperatures needed for chemical kinetic model validation. The effect of equivalence ratio on ignition delay is shown in Figure 5. The ignition delays are shown to decrease monotonically with an increase in equivalence ratio. Moreover, the change in ignition delays is larger when the equivalence ratio is increased from ϕ = 0.5 to ϕ = 1.0 at PC = 15 bar as opposed to the change when the equivalence ratio is increased from ϕ = 1.0 to ϕ = 1.5 at PC = 30 bar. In the current study, the equivalence ratio was varied by changing the fuel concentration while maintaining the oxygen concentration nearly constant (c.f., Table 1). Therefore, the change in equivalence ratio is mainly due to the change in fuel concentration. As such, the fuel concentration increases approximately by a factor of 2 when increasing the equivalence ratio from ϕ = 0.5 to ϕ = 1.0 whereas the fuel concentration changes by about factor of 1.5 when increasing the equivalence ratio from ϕ = 1.0 to ϕ = 1.5. The difference in the amount of increase in fuel concentration coupled with the nonlinear dependence of the ignition delay on fuel concentration could be the reason for the larger change (decrease) in ignition delays when the equivalence ratio is raised from ϕ = 0.5 to ϕ = 1.0 than that when increasing the equivalence ratio from ϕ = 1.0 to ϕ = 1.5, although the effect of pressure is expected to play a role as well.

Table 1. Test Conditions Investigated in the Current Study ϕ

X1‑MN (%)

XO2 (%)

XN2 (%)

PC (bar)

0.5 1.0 1.5

0.77 1.53 2.28

20.85 20.68 20.52

78.38 77.79 77.20

15, 40 15, 30, 40 30

experiments were conducted at a preheat temperature of T0 = 420.15 K. As 1-MN is prone to soot formation, cautions have to be taken during autoignition experiments, especially for fuel rich conditions. Earlier RCM work on toluene and benzene by Mittal and Sung14 demonstrated that deposition of soot on the reaction chamber could affect the autoignition characteristics of the fuel and the experimental reputability. Therefore, in the current study we regularly inspected and cleaned the reaction chamber to make sure that soot deposition (if any) did not influence the current experiments.

3. EXPERIMENTAL RESULTS Figure 3 shows the experimental pressure traces recorded at varying compressed temperatures and equivalence ratios for PC = 30 and 40 bar. Observation of the pressure traces shows no evidence of two-stage ignition behavior. In addition, no evidence of strong preignition heat release was found for the conditions tested herein. This is consistent with the shock tube experiments of Wang et al.,10 who reported that they did not observe any preignition heat release in their experiments. It is also to be noted that since the conditions shown in Figure 3 cover those with equivalence ratios ranging from fuel lean (ϕ = 0.5) to fuel rich (ϕ = 1.5) and elevated pressures of PC = 30 and 40 bar, it can be stated with confidence that 1-MN does not exhibit any significant preignition heat release across a wide range of pressures and temperatures.

Figure 3. Plot showing the pressure traces at varying compressed pressures, compressed temperatures, and equivalence ratios. C

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 4. Comparison of ignition delays of 1-MN/air mixtures at varying compressed pressures, compressed temperatures, and equivalence ratios, obtained from the current RCM experiments and the shock tube experiments of Pfahl et al.9 and Wang et al.10

the works of Mati et al.8 and Bounaceur et al.11 The model of Mati et al.8 did not include the H-abstraction and Hreplacement reactions on the aromatic rings whereas that of Bounaceur et al.11 included these channels. Since the model of Wang et al.10 considered all the possible abstraction reactions and the subsequent reactions in refs 8 and 11, the models of Mati et al.8 and Bounaceur et al.11 were not considered herein. 4.1. Ignition delay simulations. Experimental pressure traces shown in Figures 1−3 demonstrate that the instantaneous pressure in the reaction chamber drops right after the end of the compression stroke. The drop in pressure during the induction period can be attributed to the heat transfer between

4. SIMULATED RESULTS AND CHEMICAL KINETIC ANALYSES As discussed earlier, there are several chemical kinetic models of 1-MN available in the literature.8,10−12 In the current study, the models of Wang et al.10 and Narayanaswamy et al.12 were employed for simulating the experimental data and conducting the subsequent chemical kinetic analyses, as they are the latest models in the literature. The model of Narayanaswamy et al.12 comprises 158 species and 1049 reactions, while that of Wang et al.10 comprises 662 species and 3864 reactions. It is further noted that the model of Wang et al.10 was developed based on D

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 5. Plot showing the effect of equivalence ratio on ignition delay.

the hotter test mixture and the colder chamber walls. The changes in the pressure and temperature of the core gas mixture can affect the resulting ignition delay. Furthermore, the test mixture during the compression phase, even without exhibiting any significant heat release, can still lead to the buildup of the radical pool. An earlier study15 has found that it is necessary to include these thermodynamic and chemical processes during the compression and post-compression periods while simulating the RCM ignition delay. One popular way to model these processes is by conducting a “full RCM simulation”, in which a time history of volume is specified. This volume history includes both the compression phase and the post-compression period, and it is modeled using the adiabatic core hypothesis along with the corresponding nonreactive pressure trace. The nonreactive pressure trace was obtained from the experiment conducted by replacing oxygen in the oxidizer with nitrogen while keeping the fuel proportion in the mixture constant. As such, the thermal properties of the reactive and nonreactive mixtures were held identical to ensure similar compression and heat loss characteristics. The volume histories and the experimental reactive pressure traces along with a summary of ignition delay measurements are available on our Web site at http://combdiaglab.engr.uconn.edu/database/rcm-database. Typically, the ignition delay simulations were carried out using the Chemkin-Pro package16 with the volume histories and the intake conditions from experiments. Unfortunately, there were issues with convergence of the solution when full RCM simulations using Chemkin-Pro16 were attempted with the model of Wang et al.,10 while no such problem was encountered when using the model of Narayanaswamy et al.12 However, the convergence problem with the Wang et al.10 model was not encountered in “post-compression simulation”, in which the simulation excludes the compression stroke and only the post compression period is modeled with PC and TC as the initial pressure and temperature, respectively. It was further found that the convergence problem with the Wang et al.10 model in Chemkin-Pro 16 was not encountered while conducting full RCM simulations using an IdealGasReactor in Cantera v2.2.0 software.17 Hence, the full RCM simulations with the models of Wang et al.10 and Narayanaswamy et al.12 were conducted using Cantera v2.2.017 and Chemkin-Pro,16 respectively. To ensure consistency with the full RCM simulations obtained from both the software packages, Figure 6 shows a simulated result comparison using Chemkin-Pro16 and Cantera v2.2.017 with the model of Narayanaswamy et al.12

Figure 6. Plot showing the different types of full RCM simulations conducted in the current study and their comparison with the experimental result.

It is seen from Figure 6 that both the software packages essentially provide identical results. In the next sections, the experimental and simulated ignition delay results will be compared. 4.1.1. Simulations with Wang et al. model. Figure 7 shows the comparison of the experimental and simulated ignition delays obtained using the chemical kinetic model of Wang et al.10 at various test conditions. For the stoichiometric and fuel rich conditions investigated, the model of Wang et al.10 is consistently more reactive than the current experiments conducted at lower temperatures. In particular, the simulated ignition delays are approximately a factor of 2 lower than the experimental values for the conditions of ϕ = 1.0 and ϕ = 1.5. Additionally, the difference in simulated and experimental ignition delays of stoichiometric mixtures gradually increases with decreasing temperature. For the fuel lean condition (ϕ = 0.5), on the other hand, the simulated ignition delays compare reasonably well with the experiments at PC = 15 bar while the model is significantly less reactive than the experiments at PC = 40 bar. For PC = 40 bar at a similar TC value, Figure 8 compares the experimental and simulated pressure traces at stoichiometric and fuel lean conditions. For the stoichiometric condition, Figure 8(a) shows that the simulated pressure trace using the model of Wang et al.10 displays significant preignition heat release, while the experimental profile indicates otherwise. It is seen from Figure 8(b) for the fuel lean condition that the E

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 7. Comparison of the ignition delays from experiments and simulations. Filled symbols: experiments. Broken lines: simulations using the model of Wang et al.10 Solid lines: simulations using the model of Narayanaswamy et al.12

concentration profiles. Figure 9 shows such results for ϕ = 0.5 in air with initial conditions of 40 bar and 900 K. This simulation demonstrates that more than 90% of the parent fuel, 1-MN (A2CH3), is consumed by the end of the first-stage-like pressure rise and 1-MN gets converted to several intermediates. The major intermediates formed are 5-methylnaphthalene-1,4dione (A1OC6H4Obis), naphthalene-1,4-dione (A1OC6H4O), 1-naphthanol (C10H7OH), naphtalene-1carbaldehyde (C10H7HCO), and 1H,1′H-2,2′-dindene (C18H14). The chemical structures of these intermediate species are shown in Figure 9(b). From Figure 9, such high

simulated pressure history using the model of Wang et al.10 exhibits a two-stage ignition like behavior. This is somewhat surprising and unexpected for 1-MN. As shown in Figures 3 and 8, no evidence of such a distinct pressure profile was noticed in the current RCM experiments. One possible reason for this model to predict such a distinct pressure profile could be that the intermediates which are formed during the chemical induction period are not consumed, thereby delaying the hot ignition event. In order to check this hypothesis, a simple adiabatic constant volume simulation was conducted to examine the intermediate species F

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 8. Plot showing the comparison of pressure traces from experiments and simulations.

Figure 9. (a) Plot showing the evolution profiles of pressure and intermediate species predicted by the model of Wang et al.10 (b) Chemical structures of the intermediate species shown in (a). Initial conditions are P0 = 40 bar and T0 = 900 K.

concentrations of these intermediates support our hypothesis that the pressure history observed from the simulation is because of the slow oxidation of these intermediates. Further analysis was conducted to identify the reaction pathways leading to the production and consumption of the intermediate species. Reaction path analysis showed that 5methylnaphthalene-1,4-dione (A1OC6H4Obis) and naphthalene-1,4-dione (A1OC6H4O) are formed via the following two reactions of the parent fuel radicals (C6H3A1CH3 and C6H4A1CH3). C6H3A1CH3 + O2 = A1OC6H4Obis + H

(R1)

C6H4A1CH3 + O2 = A1OC6H4O + CH3

(R2)

The reactions competing with the formation of 5-methylnaphthalene-1,4-dione (A1OC6H4Obis) and naphthalene-1,4-dione (A1OC6H4O) are as follows. C6H3A1CH3 + O2 = C11H9O + O

(R3)

C6H4A1CH3 + O2 = OC6H4A1CH3 + O

(R4)

Figure 10. Reaction scheme showing the reaction channels leading to the formation of A1OC6H4Obis and A1OC6H4O in the model of Wang et al.10 “X” represents H, O, and OH.

reactions (included in the model) which lead to the formation of carbon monoxide (CO) and indenes. Moreover, R3 and R4 written in the reverse direction are the association reactions of two radicals leading to the formation of a stable species (O2 in this case) and a radical. However, reactions of such nature are not expected to proceed in the reverse direction. One possible

Figure 10 summarizes the reaction channels of R1−R4. Interestingly, R3 and R4 are also observed to proceed in the reverse direction, resulting in methyl naphthalenyl radicals (C6H3A1CH3 and C6H4A1CH3) that are predominantly consumed by R1 and R2, instead of undergoing decomposition G

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

thermodynamic properties of 5-methylnaphthalene-1,4-dione (A1OC6H4Obis) and naphthalene-1,4-dione (A1OC6H4O) in the model of Wang et al.,10 it was found that both species are assigned to have the same heat of formation as −61.06 kcal/ mol. Using RMG,20 the heats of formation of the 5methylnaphthalene-1,4-dione (A1OC6H4Obis) and naphthalene-1,4-dione (A1OC6H4O) are estimated to be −40.57 and −32.58 kcal/mol, respectively. From these checks, we are suspicious that the thermodynamic properties of some of the species in the model of Wang et al.10 might have errors. The errors in the thermodynamic properties might be affecting the equilibrium constants and the rates of the reverse reactions. The errors in the thermodynamic properties could be leading to the strange behavior observed in Figure 8. 4.1.2. Simulations with the Narayanaswamy et al. model. The performance of the model of Narayanaswamy et al.12 in predicting the ignition delays of 1-MN is also shown in Figure 7. The model of Narayanaswamy et al.,12 like that of Wang et al.,10 appears to be more reactive than experiments at the lower end of the compressed temperatures investigated in the current study. In addition, the model of Narayanaswamy et al.12 is seen to predict the ignition delays at the stoichiometric (ϕ = 1.0) condition to better accuracy than at the fuel lean (ϕ = 0.5) and fuel rich (ϕ = 1.5) conditions. It is further noticed that the difference between experimental and simulated ignition delays decreases gradually with increasing temperature up to a compressed temperature where a crossover in reactivity is observed. Beyond the crossover point, the model of Narayanaswamy et al.12 becomes less reactive than the current RCM experiments. Comparison of the model of Narayanaswamy et al.12 against the shock tube data of Wang et al.10 shown in Figure 11 also demonstrates that this model is consistently less reactive than the high temperature experiments. Therefore, the differences in reactivity between the RCM experiments and the model predictions at higher compressed temperatures accessed in the current study are consistent with that shown in Figure 11. Furthermore, the overall performance of the model of Narayanaswamy et al.12 indicates that this model does not accurately predict the global activation energy, and hence model refinements are needed. The pressure traces from experiments and simulations compared in Figure 8 show that the model of Narayanaswamy et al.12 also exhibits a strong preignition heat release for the fuel lean condition while such a strong preignition heat release is not displayed at the stoichiometric condition. Nevertheless, the pressure history simulated using the model of Narayanaswamy et al.12 is not as peculiar as seen in that predicted by the model of Wang et al.10 Based on the comparative results of Figures 7 and 8, it is clear that the model of Narayanaswamy et al.12 performs better than that of Wang et al.10 in predicting the autoignition behavior of 1-MN. Hence, the subsequent brute force sensitivity analyses were conducted using the model of Narayanaswamy et al.12 4.2. Brute force sensitivity analyses. The controlling chemistry at varying temperatures and equivalence ratios was identified using brute force sensitivity analyses. In these analyses, the pre-exponential factors of reactions were perturbed by a factor of 5.0 and 0.2 one at a time and the resulting ignition delays with the modified reaction rates were compared to capture their effects on ignition delay. The change in ignition propensity with the modification of the preexponential factor is represented by the sensitivity coefficient

scenario for such reactions to proceed in the reverse direction is when the equilibrium constants are inaccurate due to the issues with the thermodynamic properties of the species involved. As such, the thermodynamic properties of the species involved in the reactions of R1−R4 were examined. Recent quantum chemistry studies have shown that the energetics of the analogous reactions of toluene, benzene, and 1-MN derived radical species are nearly identical.18,19 Therefore, the heats of reaction for the 1-MN related reactions were compared with those from the benzene and toluene submechanisms present in the model of Wang et al.10 The choice for comparing with the reactions of benzene and toluene is because the associated chemical kinetics have been long studied and are considered to be well developed. In the following analysis, the heats of reaction (ΔHrxn) for R1−R4 are examined. First, the heats of reaction of R5 and R6, which involve H abstraction from 1-MN (A2CH3) leading to the formation of methyl naphthalenyl radicals and hydrogen molecule (H2), are compared. 1‐MN + H = C6H3A1CH3 + H 2 ΔH rxn = −1.8 kcal/mole

(R5)

1‐MN + H = C6H4A1CH3 + H 2 ΔH rxn = −1.8 kcal/mole

(R6)

The ΔHrxn values of R5 and R6 are the same as they are analogous reactions that involve H abstraction from the aromatic ring. Second, the heats of reactions for the H abstraction reactions from benzene (C6H6) and toluene (C6H5CH3) are as follows. C6H6 + H = C6H5 + H 2 ΔH rxn = 6.67 kcal/mole

(R7)

C6H5CH3 + H = C6H4CH3 + H 2 ΔH rxn = 6.67 kcal/mole

(R8)

Based on analogy, it is expected that the heats of reactions for R5−R8 should be similar. However, the ΔHrxn comparison for R5−R6 and R7−R8 shows that they differ by about 8.5 kcal/ mol. Further checks of the heats of reaction for the analogous reaction pairs (R1, R9) and (R2, R10) also exhibit significant differences, as shown below. C6H3A1CH3 + O2 = A1OC6H4Obis + H ΔH rxn = −86.00 kcal/mole

(R1)

C6H5 + O2 = OC6H4O + H ΔH rxn = −51.32 kcal/mole

(R9)

C6H4A1CH3 + O2 = A1OC6H4O + CH3 ΔH rxn = −103.82 kcal/mole

(R2)

C6H4CH3 + O2 = OC6H4O + CH3 ΔH rxn = −61.25 kcal/mole

(R10)

The difference in ΔHrxn for R1 and R9 is 34.68 kcal/mol, while the ΔHrxn values of R2 and R10 differ by 42.57 kcal/mol. Such large differences are startling. Upon further checks on the H

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

adiabatic constant volume simulations. Based on the definition of the sensitivity coefficient used here, the reactions which the accelerate ignition event upon increasing their reaction rates exhibit negative sensitivity coefficients while the reactions which retard the ignition event upon increasing their reaction rates exhibit positive sensitivity coefficients. The brute force sensitivity coefficients were computed for 450 reactions in the model of Narayanaswamy et al.12 Specifically, the reactions whose sensitivity coefficients were computed include the total aromatic chemistry described in the model, and the sensitivity coefficients of the C0−C4 reactions were not computed. The first set of brute force sensitivity analyses was conducted to study the effect of the equivalence ratio at initial conditions of P0 = 40 bar and T0 = 870 K. As demonstrated earlier, the simulated ignition delays using the base model appear to match better at the stoichiometric condition of ϕ = 1.0 than at the fuel lean condition of ϕ = 0.5. This analysis could help in identifying the reactions whose effects on ignition delay change with change in equivalence ratio. Figure 12 shows and compares the sensitivity coefficients at two different equivalence ratios, for the top 5 reactions that promote ignition and the top 5 reactions that retard ignition. The sensitivity coefficients of the reactions of 1-naphthylmethyl (A2CH2) radical with O2 and HO2 are negative, as these reactions are chain branching in nature and hence accelerate the ignition event. The Habstraction reaction from 1-MN (A2CH3) by HO2 also exhibits a negative sensitivity coefficient, as this reaction results in the buildup of hydrogen peroxide (H2O2), which is known to control ignition at low temperatures. It is noted that the reactions of naphthalen-1-yl oxidanyl (A2O) with O and O2 exhibit opposite sensitivity coefficients due to the contrasting nature of the two competing reactions. Reaction of naphthalen1-yl oxidanyl (A2O) with O2 would produce OH radical at the expense of O2, while the reaction of naphthalen-1-yl oxidanyl (A2O) with O radical produces an H atom. H atom at the current temperature of 870 K gets converted to HO2 radical, which is less reactive than OH radical. This difference in

Figure 11. Plot showing the performance of the model of Narayanaswamy et al.12 at the conditions investigated by Wang et al.10

defined as

(ln ) (ln ), τ5.0 τ0.2

5.0 0.2

where τ0.2 and τ5.0 are the

simulated ignition delays using the reaction rate multiplied by 0.2 and 5.0, respectively. Both τ0.2 and τ5.0 were obtained from

Figure 12. Comparison of brute force sensitivity coefficients of selected reactions at varying equivalence ratios for initial conditions of 40 bar and 870 K. I

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

computed and compared at initial temperatures of 870 and 1200 K, as they are the typical conditions investigated in the current RCM experiments and the shock tube study of Wang et al.,10 respectively. This analysis aims to identify if there exists any reaction(s) whose effect on ignition delay changes characteristically from promoting to retarding or vice versa upon change in temperature. The comparison of sensitivity coefficients computed at the two initial temperatures for those selected, important reactions is shown in Figure 14. From Figure 14, the only reaction whose sensitivity coefficients change sign with different initial temperatures is the reaction between 1-MN and O radical. In the model of Narayanaswamy et al.,12 the reaction system between 1-MN and O has been described as a lumped scheme leading to the formation of carbon monoxide, methyl radical, and indenyl radical. The reaction rate of this lumped scheme used in the model of Narayanaswamy et al.12 is analogous to the rate of the reaction between benzene and O radical leading to the formation of phenyl oxidanyl (C6H5O) and H radicals that was taken from the benzene submodel of Blanquart et al.21 Taatjes et al.22 have recently conducted a study focusing on the reaction channels of benzene with O radical and proposed the rates of all the possible reaction channels. Figure 15 shows the comparison of the reaction rate used in the model of Narayanaswamy et al.12 and the high pressure limit rate taken from Taatjes et al.22 It is seen that the reaction rate proposed by Taatjes et al.22 is about a factor of 5 lower than the rate used in the model of Narayanaswamy et al.12 at 830 K and the difference in reaction rates drops to a factor of 2 at 1200 K. As this particular reaction exhibits a positive (negative) sensitivity coefficient at 870 K (1200 K), using the rate from Taatjes et al.22 would make the model relatively more (less) reactive at low (high) temperatures and hence may not be a solution to resolve the current discrepancy between experiments and simulations. Unfortunately, the sensitivity analyses conducted here were not able to identify the reactions whose rates can be modified to improve the performance of the base model. One possible reason for the differences observed in the experimental and simulated ignition delays can be due to the absence of some important reaction paths in the model of Narayanaswamy et al.12 4.3. Possible missing reaction paths. The literature studies of refs 18 and 22−25 were surveyed to help identify some of the possible missing reaction pathways. In the study of Taatjes et al.22 on the addition reaction channels of benzene (C6H6) and O radical, three major reaction pathways, R11− R13, were noted.

reactivities of the resulting OH and HO2 is the reason why the reactions of naphthalen-1-yl oxidanyl (A2O) with O2 and O exhibit contrasting sensitivity on ignition delay. In addition, the chain initiation reaction between 1-MN (A2CH3) and O2 leading to the formation of 1-naphthylmethyl (A2CH2) radical and HO2 radical retards ignition, because this reaction was observed to proceed in the reverse direction, thus competing with the reaction leading to formation of naphtalene-1carbaldehyde (A2CHO) and OH radical, which accelerates ignition, as indicated by its negative sensitivity coefficient. The substitution reaction of OH with 1-MN leading to the formation of naphthalen-1-ol (A2OH) and CH3 exhibits a positive sensitivity coefficient, as this reaction competes with the reaction of OH with 1-MN leading to the formation of 1naphthylmethyl (A2CH2) radical, scavenges OH radical, and produces naphthalen-1-ol (A2OH), which is less reactive. Furthermore, Figure 12 shows that the sensitivity coefficients of most of the reactions do not change significantly with equivalence ratio. One exception is the reaction between 1Hinden-1-one (C9H6O) and H, leading to the formation of 2ethenylphenyl (A1C2H3*) radical and CO. Since the sensitivity coefficient of this reaction is negative, reducing the rate of this reaction could help in slowing down the reactivity of the base model. A computational experiment was then conducted by reducing the reaction rate of this reaction by a factor of 2 to illustrate how the pressure trace is affected. Figure 13 shows the comparison of the pressure traces obtained from

Figure 13. Plot showing the effect of decreasing the rate of the reaction C9H6O + H = A1C2H3* + CO by a factor of 2 on pressure evolution.

simulations using the base and modified models. Reducing the reaction rate of this reaction is seen to cause the pressure profile to exhibit a two-stage like behavior, thereby indicating that reducing the rate might not be an appropriate way to improve the base model. The reason for this reaction to lead to such a distinct effect on pressure history is that in the model of Narayanaswamy et al.12 1H-inden-1-one (C9H6O) acts as a bottleneck for reactivity. Namely, 1H-inden-1-one (C9H6O) behaves as a relatively stable, nonreactive species limiting the breakdown of the parent fuel (1-MN) into small fragments and thus hindering ignition. This analysis shows that the submodel of 1H-inden-1-one (C9H6O) in the model of Narayanaswamy et al.12 needs further improvements. The second set of brute force sensitivity analyses was conducted to study the change in sensitive reactions with change in initial temperature, for stoichiometric mixtures with an initial pressure of 40 bar. The sensitivity coefficients were

C6H6 + O = C6H5O + H

(R11)

C6H6 + O = C5H6 + CO

(R12)

C6H6 + O = C6H5OH

(R13)

R11−R13 lead to the formation of phenyloxidanyl (C6H5O), cyclopenta-2,4-dien-1-yl radical (C5H6), and phenol (C6H5OH). Nguyen et al.23 also stated that R11 and R13 are the important channels for the addition reaction system of C6H6 + O. Both the studies of Taatjes et al.22 and Nguyen et al.23 suggested that R13 is important at low temperatures. The model of Narayanaswamy et al.12 included the lumped reaction scheme, which is analogous to R11 but does not include the reactions analogous to R12 and R13. One reason that Narayanaswamy et al.12 did not add the reactions analogous J

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 14. Comparison of brute force sensitivity coefficients of selected reactions for stoichiometric 1-MN/air mixtures with an initial pressure of 40 bar at varying initial temperatures.

Figure 15. Plot showing the rate comparison for the reaction of C6H6 + O = C6H5O + H taken from Blanquart et al.21 and Taatjes et al.22

Figure 16. Important reactions of 1-MN + OH system. Reactions with broken arrows indicate the pathways which are missing in the model of Narayanaswamy et al.12

to R13 is because Shaddix et al.6,7 did not observe any methyl substituted naphthalen-1-ols in their flow reactor experiments at 1170 K. As stated by Narayanaswamy et al.,12 there could be a possibility that products formed from the reactions analogous to R13 could be short-lived species at conditions relevant to the experiments of Shaddix et al.6,7 However, for the low-tointermediate conditions investigated in the current RCM study, the methyl substituted naphthalen-1-ols might not be shortlived species and potentially could also act as a radical sink. Therefore, it might be necessary to include the reaction channels analogous to R13. In addition, the H-abstraction reactions from the aromatic ring were not included in the model of Narayanaswamy et al.12 The different types of H-abstraction reactions are shown in Figure 16, using the 1-MN + OH system as an example. The reactions of 1-MN and OH account for the major consumption of 1-MN, while the H-abstraction from the methyl site by OH radical is the only type of H-abstraction reactions considered in the model of Narayanaswamy et al.12 In ref 12, the rate for H-

abstraction by OH radical from the methyl site was taken from Seta et al.24 The abstraction rates from Seta et al.24 suggest that, at temperature 870 K, about 71.6% of the flux of the Habstraction reactions is from the abstraction from the methyl site, while 28.4% of the flux is through the H abstractions from the aromatic ring. As such, the inclusion of the H abstraction from the aromatic ring is expected to reduce the amount of 1naphthylmethyl (A2CH2) radical produced from 1-MN. The sensitivity analyses conducted in the current study showed that the reactions of 1-naphthylmethyl (A2CH2) radical with O2 and HO2 are the two major reactions that control the reactivity of 1-MN. In addition, the subsequent reactions for the species formed from the H abstraction from the aromatic ring are not chain branching at low temperatures. Hence, the inclusion of all possible H-abstraction channels shown in Figure 16 can potentially lower the reactivity of 1-MN at low temperatures. Another important reaction pathway that is missing in the base model is a channel unique to the reactions between 1naphthylmethyl (A2CH2) radical and O2 recently observed by K

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Oguchi and Murakami,19 as shown in Figure 17. Oguchi and Murakami19 found that the reactions between 1-naphthylmeth-

respectively. However, the model of Narayanaswamy et al.12 only used the reaction rate corresponding to the temperature range 800−2000 K. As a consequence, the model of Narayanaswamy et al.12 overestimated the reaction rate of the particular reaction by factors of 13 and 4 at 600 and 700 K, respectively. In closing this section, it is noted that the chemical kinetic analyses and the literature survey conducted herein helped in identifying the possible reasons for the differences observed between experiments and simulations. In addition, the present analyses highlighted the need for a comprehensive chemical kinetic model of 1-MN that would require information on the thermodynamic properties of the intermediate species and the reaction rate rules. The present new data sets represent validation targets for the development of predictive chemical kinetic models which will necessitate more accurate thermodynamic properties and rate constants. Such a model development is very broad in its scope, requiring a comprehensive and extensive review of new thermochemistry data and further experimental and computational studies for full understanding of the underlying reaction pathways and the associated rate constants.

Figure 17. Important reactions of the 1-naphthylmethyl (A2CH2) + O2 system taken from Oguchi and Murakami.19 The reaction shown by the broken arrow indicates the pathway which is missing in the model of Narayanaswamy et al.12

5. CONCLUSIONS Autoignition of 1-MN has been studied at low-to-intermediate temperatures and elevated pressures using an RCM. The RCM experiments were conducted to demonstrate the effects of temperature, equivalence ratio, and pressure on ignition delays. The ignition delays measured in the current study were observed to complement well the literature shock tube data at higher temperatures. For the conditions investigated here, the experimental pressure traces showed that 1-MN behaves like a single-stage ignition fuel and no strong preignition heat release was observed. In addition, the experimental ignition delays were used to assess the performance of two latest chemical kinetic models of 1-MN. Of the two models, the model of Narayanaswamy et al.12 appears to do a better job in estimating the measured ignition delays and pressure histories. Since the pressure history predicted by the model of Wang et al.10 exhibited a two-stage ignition-like behavior which was not observed in the RCM experiments, an analysis was conducted to show that the thermodynamics properties of some of the intermediate species in the model need to be corrected. The ignition delays predicted by the model of Narayanaswamy et al.12 exhibited significant differences at the lower end of the compressed temperatures investigated in the current study, with discrepancies being about a factor of 1.5−2 lower than experimental values. Moreover, the model of Narayanaswamy et al.12 did show a strong preignition heat release at fuel lean conditions, which was not observed from the experimental pressure traces. Further chemical kinetic analyses were conducted to show that some of the important reaction channels are missing in the model of Narayanaswamy et al.12 The inclusion of these missing channels identified here could potentially improve the predictability of the 1-MN model.

yl (A2CH2) radical and O2 could lead to the formation of 2Hnaphtho[1,8-bc]furan (cyc-NaphCH2O) in addition to the channels leading to the formation of naphtalene-1-carbaldehyde (A2CHO) + OH and naphthalen-1-yl oxidanyl (A2O) + formaldehyde (CH2O). It is noted that the pathway leading to naphtalene-1-carbaldehyde (A2CHO) + OH accounts for the major consumption of 1-naphthylmethyl (A2CH2) radical, while the channel forming and naphthalen-1-yl oxidanyl (A2O) + CH2O are insignificant therein. Since the channels leading to the formation of naphtalene-1-carbaldehyde (A2CHO) and naphthalen-1-yl oxidanyl (A2O) radicals are identical to those observed in the benzyl + O2 system,25 their reaction rates in the model of Narayanaswamy et al.12 were taken from ref 25. Oguchi and Murakami19 estimated that at 10 atm and in the temperature range 700−1300 K, the branching ratio of the channel leading to the formation of 2H-naphtho[1,8-bc]furan (cyc-NaphCH2O) is about 0.7 while the branching ratio leading to the formation of naphtalene-1-carbaldehyde (A2CHO) and naphthalen-1-yl oxidanyl (A2O) is around 0.15, indicating that the missing channel could be dominant. As the reaction between 1-naphthylmethyl (A2CH2) radical and O2 leading to naphtalene-1-carbaldehyde (A2CHO) + OH was found to be the most sensitive reaction in the present brute force sensitivity results, addition of the channel leading to the formation of 2H-naphtho[1,8-bc]furan (cyc-NaphCH2O) might be important. Finally, the other missing channel is the addition reaction between 1-naphthylmethyl (A2CH2) and HO2 radicals leading to the formation of (naphthalen-1-yl)methaneperoxol (A2CH2OOH) adduct, in addition to that leading to 1methylnapthoxyl radical (A2CH2O) + OH. The formation of adduct is expected to be important at temperatures below 700 K and compete with the formation of 1-methylnapthoxyl radical (A2CH2O) + OH, based on the work of da Silva and Bozzelli.26 Furthermore, da Silva and Bozzelli26 noticed that the reaction rate leading to the formation of 1-methylnapthoxyl radical (A2CH2O) + OH shows a complex temperature dependence and proposed two sets of rate parameters for the temperature ranges of 300−800 K and 800−2000 K,

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (860) 486-2492. Fax: (860) 486-5088. ORCID

Goutham Kukkadapu: 0000-0003-0240-4028 L

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Science Foundation under Grant No. CBET-1402231. GK was also supported by the Lawrence Livermore National Laboratory via Standard Research Subcontract No. B617843.

■

REFERENCES

(1) Gieleciak, R.; Fairbridge, C. Detailed hydrocarbon analysis of FACE diesels using comprehensive two-dimensional gas chromatography. http:// www.crcao.org/publications/advancedVehiclesFuelsLubricants/ FACE/index.html. (2) Mueller, C. J.; Cannella, W. J.; Bruno, T. J.; Bunting, B.; Dettman, H. D.; Franz, J. A.; Huber, M. L.; Natarajan, M.; Pitz, W. J.; Ratcliff, M. A.; Wright, K. Energy Fuels 2012, 26, 3284−3303. (3) Mueller, C. J.; Cannella, W. J.; Bays, T.; Bruno, T. J.; DeFabio, K.; Dettman, H. D.; Gieleciak, R.; Huber, M. L.; Kweon, C.-B.; McConnell, S. M.; Pitz, W. J.; Ratcliff, M. A. Energy Fuels 2016, 30, 1445−1461. (4) Anand, K.; Ra, Y.; Reitz, R. D.; Bunting, B. Energy Fuels 2011, 25, 1474−1484. (5) Farrell, J. T.; Cernansky, N. P.; Dryer, F. L.; Law, C. K.; Friend, D. G.; Hergart, C. A.; David, R. M.; Patel, A. K.; Mueller, C. J.; Pitsch, H. SAE 2007-01-0201. (6) Shaddix, C. B.; Brezinsky, K.; Glassman, I. Symp. (Int.) Combust., [Proc.] 1992, 24, 683−690. (7) Shaddix, C. B.; Brezinsky, K.; Glassman, I. Combust. Flame 1997, 108, 139−157. (8) Mati, K.; Ristori, A.; Pengloan, G.; Dagaut, P. Combust. Sci. Technol. 2007, 179, 1261−1285. (9) Pfahl, U.; Fieweger, K.; Adomeit, G. Symp. (Int.) Combust., [Proc.] 1996, 26, 781−789. (10) Wang, H.; Warner, S. J.; Oehlschlaeger, M. A.; Bounaceur, R.; Biet, J.; Glaude, P. A.; Battin-Leclerrc, F. Combust. Flame 2010, 157, 1976−1988. (11) Bounaceur, R.; Glaude, P. A.; Fournet, R.; Battin-Leclerc, F. Int. J. Vehicle Des. 2007, 44, 124−142. (12) Narayanaswamy, K.; Blanquart, G.; Pitsch, H. Combust. Flame 2010, 157, 1879−1898. (13) Das, A. K.; Sung, C. J.; Zhang, Y.; Mittal, G. Int. J. Hydrogen Energy 2012, 37, 6901−6911. (14) Mittal, G.; Sung, C. J. Combust. Flame 2007, 150, 355−368. (15) Mittal, G.; Chaos, M.; Sung, C. J.; Dryer, F. L. Fuel Process. Technol. 2008, 89, 1244−1254. (16) CHEMKIN-PRO; Reaction Design: San Diego, 2013. (17) Goodwin, D. G.; Moffat, H. K.; Speth, R. L. Cantera: an objectoriented software toolkit for chemical kinetics, thermodynamics, and transport processes; 2015. http://www.cantera.org. (18) Zhou, C. W.; Kislov, V. V.; Mebel, A. M. J. Phys. Chem. A 2012, 116, 1571−1585. (19) Oguchi, T.; Murakami, Y. Chem. Phys. Lett. 2013, 585, 27−32. (20) RMG − Reaction Mechanism Generator, http://rmg.mit.edu/. (21) Blanquart, G.; Pepiot-Desjardins, P.; Pitsch, H. Combust. Flame 2009, 156, 588−607. (22) Taatjes, C. A.; Osborn, D. L.; Selby, T. M.; Meloni, G.; Trevitt, A. J.; Epifanovsky, E.; Krylov, A. I.; Sirjean, B.; Dames, E.; Wang, H. J. Phys. Chem. A 2010, 114, 3355−3370. (23) Nguyen, T. L.; Peeters, J.; Vereecken, L. J. Phys. Chem. A 2007, 111, 3836−3849. (24) Seta, T.; Nakajima, M.; Miyoshi, A. J. Phys. Chem. A 2006, 110, 5081−5090. (25) Murakami, Y.; Oguchi, T.; Hashimoto, K.; Nosaka, Y. J. Phys. Chem. A 2007, 111, 13200−13208. (26) da Silva, G.; Bozzelli, J. W. Proc. Combust. Inst. 2009, 32, 287− 294.

M

DOI: 10.1021/acs.energyfuels.6b01628 Energy Fuels XXXX, XXX, XXX−XXX