Biologically derived diesel fuel and NO formation

Combustion and Flame 158 (2011) 2302–2313

Contents lists available at ScienceDirect

Combustion and Flame j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e

Biologically derived diesel fuel and NO formation Part 2: Model development and extended validation S. Garner a, T. Dubois b, C. Togbe b, N. Chaumeix b, P. Dagaut b, K. Brezinsky a,⇑ a b

Department of Mechanical and Industrial Engineering, University of Illinois at Chicago, Chicago, IL 60607, USA Centre National de la Recherche Scientifique, ICARE, 45071 Orleans Cedex 2, France

a r t i c l e

i n f o

Article history: Received 18 February 2011 Received in revised form 14 June 2011 Accepted 15 June 2011 Available online 30 July 2011 Keywords: Chemical kinetics Biodiesel NO Shock tube High pressure

a b s t r a c t Part 1 of this two part series presented a chemical kinetic model for the simulation of high pressure shock tube pyrolysis and oxidation data of two representative biodiesel surrogate components and the application of this model for predicting prompt NO at practical diesel combustion conditions. The present work discusses in greater detail the model’s development, structure, and rate parameters as well as expands the model’s validation range to include complementary 10 atm jet stirred reactor (JSR) oxidation experiments conducted at lower temperatures (550–1200 K) and longer reaction times of 0.7 s. In addition, shock tube ignition delay measurements of 1-heptene and 1,6-heptadiene, analogs of the hydrocarbon side chains of the methyl esters, have also been performed and are presented to further constrain the model. Ó 2011 Published by Elsevier Inc. on behalf of The Combustion Institute.

1. Introduction With stricter emissions regulations being implemented, there is a need to reduce pollutant emissions from diesel engines burning petro-diesel. Biologically derived diesel or biodiesel is known to produce fewer pollutants than petro-diesel and is now commercially blended with petro-diesel in small amounts (generally less than 5% by volume). Biodiesel appears to have only one drawback in its emission characteristics in that it produces increased NOx emissions [1–4] when compared to conventional diesel. As such two representative biodiesel surrogate components, methyl octanoate and methyl trans-2-octenoate (seen in Fig. 1) were studied in detail in Part 1 [5] to determine the link between fuel structure, acetylene, and prompt NO formation. Although the trans isomer of methyl octenoate is not considered to be a bioderived ester due to the fact that unsaturated bioderived esters are generally cis, it can nevertheless be used as a representative species for the study of the chemistry of the unsaturated methyl esters available from biological sources since the cis–trans conformation is not playing a role in the chemistry being studied. Similarly, this readily available unsaturated methyl ester, with the double bond at the second position, can be used to examine the basic chemistry of unsaturated methyl esters and their link to acetylene and prompt NO formation. Part 1 [5] of this two part series focused primarily on modeling the increased acetylene formation that was experimentally seen ⇑ Corresponding author. E-mail address: [email protected] (K. Brezinsky).

from the unsaturated methyl ester fuel compared to the saturated fuel and studying acetylene’s link to prompt NO formation. In addition, a determination of the role of the methyl ester group on the overall acetylene formation and therefore prompt NO was also discussed. The prompt NO predictions and subsequent discussion in Part 1 [5] was accomplished by the application of the detailed chemical kinetic model developed in this work. To preserve the clarity of the prompt NO discussion in Part 1 [5] the developed chemical kinetic model itself was not discussed in great detail. This article, Part 2, focuses on the development, structure, and rate parameters adopted for the chemical kinetic model presented in Part 1 and the extension of the validation range. The following section will give a brief background of the work presented in Part 1 [5]. 2. Part 1: summary To examine the possible chemical sources of the increased NOx that is seen during the combustion of biodiesel fuels, when compared to conventional petroleum based diesel fuel, experiments have been performed at relevant combustion pressures of 27 atm and 53 atm, and reactions times, 1.65 ms. Representative saturated and unsaturated components of biodiesel surrogates, methyl octanoate (C9:0) and methyl trans-2-octenoate (C9:1), were studied in a high pressure single pulse shock tube under the rich oxidation conditions, U  1 to U  3, that are present during diesel combustion [6]. It was hypothesized, and seen experimentally, that increased amounts of acetylene, C2H2, were formed from the unsaturated methyl ester compared to the saturated methyl ester.

0010-2180/$ - see front matter Ó 2011 Published by Elsevier Inc. on behalf of The Combustion Institute. doi:10.1016/j.combustflame.2011.06.011

2303

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

CH3

CH3

O

H3C

O

O

H3C

O

C7 Alkyl Side Chain methyl octanoate (C9H18O2)

methyl trans-2-octenoate (C9H16O2) Fig. 1. Surrogate fatty acid methyl esters.

Acetylene was hypothesized to be, and used as, a marker to predict prompt NO formation due to its known coupling to prompt NO through methylidyne (CH) production under rich oxidation conditions (Reactions (1) and (2)). The experimental shock tube data of stable intermediate species mole fractions as a function of temperature, pressure, and reaction time were used to develop and validate a chemical kinetic model for the rich oxidation and pyrolysis of methyl octanoate (C9:0) and methyl trans-2-octenoate (C9:1). Following the model’s validation against the experimental shock tube data it was coupled to an existing prompt NO mechanism from Konnov [7]. Using the model in a predictive capacity it was seen that over the specific temperature range of 1050–1450 K larger amounts of NO were predicted from the unsaturated fuel, methyl octenoate (C9:1) with a maximum difference in the mole fraction predictions to be approximately an order of magnitude at 1200 K. These predictions therefore support the hypothesis within the literature that a portion of the increased NOx seen from the combustion of biologically derived diesel fuel is due to chemical effects [8,9].

C2 H2 þ O ¼ CH2 þ CO

ð1Þ

H þ CH2 ¼ CH þ H2

ð2Þ

The authors are aware of studies where some engine test have yielded little change (no change, slight increase or slight decrease) in NOx formation from engines fueled with biodiesel fuel. It is extremely difficult to draw conclusion from a wide range of engine studies where the experimental test engines, EGR rates, injection systems, and combustion cylinder geometries do not remain constant. The authors acknowledge that under certain conditions with certain test engines it is possible to have little change between the NOx emissions from the same engine fueled with petrodiesel and biodiesel but the majority of the engine studies within the literature illustrate a link between increased NOx formation and the use of biodiesel fuels. This work contributes to the hypothesis that chemical sources of NOx (prompt NOx) contribute to the overall increased NOx formation from biodiesel fueled systems that have been seen experimentally [1–4]. In particular, this manuscript reports (along with Part 1 [5] of this series) results that link the fuel structure, its effect on acetylene formation, and by implication the amount of experimentally observed NO formed by the prompt NO mechanism. In this sense, the work does indeed explain, perhaps only in part, but nevertheless does explain why different biodiesel fuels may lead to increased levels of emitted NO. There exists, within the available literature, kinetic models for the simulation and prediction of fatty acid methyl ester oxidation experiments such as the methyl decanoate mechanism of Herbinet et al. [34]. The methyl decanoate model does in fact contain some reaction steps for the consumption of methyl octanoate as well as the species methyl trans-2-octenoate, though very little reaction steps involving this species. The model of Herbinet et al. contains 1880 species and 8555 reactions and it is possible that this model

may provide reasonable predictions but the model itself is prohibitively large and impractical from an application standpoint. With this in mind, the kinetic model proposed in this manuscript (Part 2) as well as Part 1 [5], is significantly smaller with only 1707 reactions and therefore is much better adapted to the modeling of applications. Furthermore, the proposed model encompasses both practical combustion conditions for compression ignition engines (Part 1 – HPST) as well as lower temperature/longer reaction time systems (JSR – Part 2) for saturated and unsaturated C8 methyl esters.

3. Experimental setup 3.1. High pressure single pulse shock tube (HPST) The experimental condition, setup, analyses, errors, and data related to the HPST are extensively discussed in Part 1 [5] of this series but for completeness a brief description of the HPST is as follows. The HPST has a driver section 6000 long with a 200 bore and (for this study) a driven section 10100 long with a 100 bore and the two sections are separated by scored aluminum diaphragms. The shock tube operates in a single pulse fashion with a dump tank situated close to the diaphragm section to quench the transmitted portion of the reflected shock wave. Tuning of the HPST is made possible by varying the length of the driver section via the use of solid brass plugs in order to obtain optimal reaction quenching rates of approximately 3.6  105 K/s. The incident shock velocities were calculated using the response from six, axial spaced, sidewall mounted pressure transducers (PCB model 113A21). The reaction temperature, T5, for each shock is calculated using an external chemical thermometer which correlates the measured shock velocity (uncertainty 6 1%) with the gas temperature within the reaction zone [10]. These shock velocities have been calibrated, external to the experimental sets, by measuring the extent of the unimolecular decomposition of 1,1,1-trifluoroethane which has a well established rate constant from Tranter et al. [10]. Previous studies into the effects of the quenching time and pressure fluctuations during the experiment have found these effects to be insignificant with respect to the reaction temperature [10,11]. Contrary to shock tube ignition delay studies, where time to the onset of reaction is important and influenced by even small irregularities in the pressure/temperature profiles, the UIC high pressure shock tube behaves, and is successfully modeled as, a closed batch reactor which undergoes an instantaneous pressure and temperature rise (caused by the passage of the reflected shock wave). Following this instantaneous temperature and pressure rise there exist a well defined reaction time, during which the reactions occur and pressure remains effectively constant. Any effects of a pressure fluctuation on the results from the way the species oriented UIC single

2304

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

pulse shock tube operates have been evaluated and discussed in earlier publications [10,11]. The uncertainty in the defined reaction temperature is estimated to be no more than ±2% over the temperature range spanning these experimental sets. The reaction pressure, P5, is found to be the average peak pressure as recorded from the end wall pressure transducer and the reaction time is empirically taken to be the time difference between the reflection of the incident shock wave and when the P5 (reflected shock pressure) pressure drops to 80% of its peak value [11]. The uncertainty of the time measurement is estimated to be no more than 10% of the measured value. Boundary layer formation is known to be reduced at high pressures such as those used in this study [12,13]. Nevertheless to insure that any effects of the boundary layer are minimized, gas samples are withdrawn from the center core gas and the sampling time has been experimentally determined to verify that core gas is sampled from very close to the endwall of the driven section. 3.2. JSR A complete description of the JSR experimental setup utilized for this study has been described in earlier publications [14,15]. The JSR consisted of a small fused-silica sphere (to minimize wall catalytic reactions) of 4 cm diameter (volume of 39 cm3), equipped with four nozzles of 1 mm i.d. for the admission of the gases which achieve the stirring. A nitrogen flow of 50 L/h was used to dilute the fuel. The nitrogen is admitted into the mass flow controller (MFC) at room temperature and a pressure equal to the reactor operating pressure +5 bars. Between the MFC and the reactor, the pressure drops to the reactor pressure (10 bar) while the temperature gradually increases to the temperature inside the reactor at the mixing region. As before [14,16–18], all the gases (provided by Air Liquide) were preheated before injection to minimize temperature gradients inside the reactor (<2 K/cm). A regulated heating wire of 1.5 kW maintained the temperature of the reactor at the desired operating temperature. Nitrogen (<50 ppm of O2; <1000 ppm of Ar; <5 ppm of H2) was used to dilute the reactants that were mixed prior to the entrance of the injectors. High-purity oxygen (99.995% pure) was used in these experiments. The fuels (methyl trans-2-octenoate P96% pure from Sigma–Aldrich and methyl octanoate P99%, verified using GC/MS) were sonically degassed before use. A liquid pump (Shimadzu LC10 AD VP) with online degasser (Shimadzu DGU-20 A3) was used to deliver the fuel to an in-house constructed atomizer-vaporizer assembly maintained at 200 °C. For each experiment, a good thermal homogeneity (gradients of 1 K/cm) along the vertical axis of the reactor was observed by thermocouple measurements (0.1 mm Pt–Pt/Rh 10% located inside a thin-wall silica tube). The reacting mixtures were sampled using a fused-silica low-pressure sonic probe (4–6 kPa, 3 mm o.d., 2 mm i.d.). The samples were taken at steady temperature and a residence time of 0.7 s. They were analyzed offline by gas chromatography after collection and storage in 1 L Pyrex bulbs. Online Fourier transform infrared (FTIR) analyses of the reacting gases were also performed by connecting the sampling probe to a temperature controlled gas cell (140 °C, 10 m path length) via a Teflon heated line (210 °C). The sample pressure in the cell was 0.2 bar and a 0.5 cm1 resolution was used. High vapor-pressure species and permanent gases were analyzed off-line whereas low vapor pressure compounds were analyzed online. The experiments were carried out at steady state and constant mean residence time, with the reactants continually flowing into the reactor while the reactor temperature was varied stepwise. A high degree of dilution was used, reducing temperature gradients and heat release. Gas chromatographs equipped with fused-silica capillary columns (Molecular Sieve-5A, DB-624, Plot Al2O3/KCl, Carboplot-

P7), a thermal conductivity detector (TCD), and a flame ionization detector (FID) were used for measuring the concentrations of stable species. Compound identifications were made through gas chromatography/mass spectrometry (GC/MS) analyses of the samples. A quadrupole mass spectrometer (GC/MS Varian 1200) operating in electron impact ionization mode (70 eV) was used. Good repeatability of the measurements and good carbon balance (100 ± 10%) were obtained in the present experiments (data tables included as supplemental material). 3.3. Shock tube ignition delay Experiments were carried out behind reflected shock waves in a stainless steel shock tube with a 2 m long driver section and a 5.15 m driven section with a 52.4 mm i.d. Both of the two shock tube portions were evacuated using two primary vacuum pumps. The shock velocity was measured via four axial pressure transducers equally spaced by 150 mm, flush mounted into the inner surface of the tube, the last one being positioned 15 mm from the end-wall. The last section has been blackened in order to reduce light reflection from the walls. Two opposite fused silica window (9 mm optical diameter and 6 mm thickness) are mounted across, the first onea narrow band filter centered at 306 nm (characteristic of OH emission) and equipped with a UV-sensitive photomultiplier tube 1P28 HAMAMATSU, and the second one across aon,ther narrow band filter centerd at 431 nm (characteristic of CH emission) and equipped with a UV sensitive photomultiplier R928 HAMAMATSU. Both pressure and emission signals are transferred and registered on a numerical oscilloscope. Reflected shock conditions were calculated from a standard procedure [19]. The ignition delay time is defined as the time between the passage of the reflected shock wave, indicated by pressure jump, monitored by a pressure transducer, and 50% of the maximum OH emission signal at 306 nm. In case of less diluted mixtures, the OH emission corresponds to the run-away of the explosive reaction detected by a small pressure signal increase. The error in the temperature is estimated to be less than 1% while that of the pressure is 1.3%. With regards to the error in the auto-ignition delay times, the estimated error depends on the temperature range and is the highest on the high temperature side, varying between 2% and 14%. The reactive gas mixtures were prepared using 1-heptene (Aldrich 99+%), 1,6-heptadiene (Aldrich 99+%), oxygen (Air liquide 99,999%), and argon (Air liquide 99,999%) in a 17 L stainless steel reservoir equipped with magnetic fans. The liquid hydrocarbon was degassed several times before the mixtures were prepared. The mixtures were mixed 1–2 h to ensure homogeneous composition prior to experiments being performed. The whole setup (shock tube, tubing and reservoir) has been heated up to an initial temperature of 90 °C in order to avoid any condensation or adsorption on the walls. 4. C8 methyl ester model development, structure, and rate parameter adoption rational The model presented in Part 1 [5] and discussed here in detail consists of 1707 reactions and 290 species. The model was developed in a hierarchical manner using the C7 hydrocarbon pyrolysis model of Garner et al. [20] as the foundation upon which the methyl ester sub mechanisms were written by analogy using the mechanisms of Fisher et al. [21] and Gail et al. [22]. Species naming conventions were also adopted from the nomenclature of Fisher et al. [21]. Simulations were performed using Chemkin 3.6 [23] and Chemkin 4.1 [24] while employing the Senkin subroutine for the modeling of the shock tube experiments and Aurora for the modeling of the JSR (CHEMKIN v3.6 only) with constant pressure and adiabatic assumptions. For this software package, the reaction

2305

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

rates are computed using the modified Arrhenius equation, k = ATnexp(Ea/RT) and the thermochemical data is in the form of the NASA polynomials. Much of the thermochemical data (included in the tradition NASA format) was taken from existing large methyl ester mechanisms [33–35] with the remainder of the thermochemical species data estimated via the group additivity methods put forth by Benson [25] employed by the software program THERGAS [26]. The complete mechanism, including references and thermochemistry, has been included as a supplementary file. Example reactions and rate parameters for each reaction class (initiation, hydrogen abstraction, isomerization, and b-scission) described below can be seen in Table 1. Some of the initiation reactions presented in Table 1 are presented in the reverse direction (e.g. radical recombination reactions) but the rate constant parameters presented in Table 1 are presented for the forward reaction defined as the reaction progressing from left to right.

m

O

5

7 8

1

3 4

6

O

2

Fig. 2. Methyl octanoate with carbons numbers associated with the nomenclature.

m O

3 4

1 2

O

Fig. 3. Methyl butanoate with carbons numbers associated with the nomenclature from Fisher et al. [21].

4.1. Initiation reactions For the unimolecular reactions of C–H bond breaking for methyl octanoate and for methyl trans-2-octenoate, the reactions were written as a recombination of the fuel radicals and hydrogen atoms with the rate constants assumed to be at their high pressure limits (within the range of 1E13–1E14) as proposed by Fisher et al. [21]. Although, the model proposed by Fisher et al. [21] is for methyl butanoate, a saturated C4 fatty acid methyl ester the rate parameters for C–H bond braking in the alkyl side chain and on the methoxy carbon of the saturated C8 fatty acid methyl ester were assumed to be the same as their analogous C4 methyl ester reaction and were adopted from Fisher et al. [21] in the following manner. The rate parameters for the C–H bond breaking on the terminal carbons (carbon number 8 for methyl octanoate and carbon number 4 for methyl butanoate – see Figs. 2 and 3 for the carbon numbering) and the second carbon atoms from both fuels were assumed to be the same. The central carbon atoms of methyl octanoate (7–3) are assumed to have the same C–H bond breaking rate parameters as the third carbon of methyl butanoate. C–H breaking on the methoxy carbon for both fuels were also assumed to be the same. The rational for this pattern of rate parameter adoption

Table 1 Reaction examples and rate parameter adoption for each reaction class. Units in cm3, s, mol, cal. Note: The naming conventions for the methyl ester species have been adopted from Fisher et al. [21]. The letter D denotes a double bond between carbon n and n + 1 while J denotes the radical position. The carbon atoms are numbered with the first carbon being the carbonyl carbon. Example: MOCT2D8J has an eight carbon chain with a double bond between carbon 2 and carbon 3 and a radical site on carbon 8. Reaction

A

n

Ea

Ref.

Initiation reaction examples MOCT7J + H = MOCT MOCT = MPEN5J + nC3H7 MOCT2D7J + H = MOCT2D MB2D4J + pC4H9 = MOCT2D MOCT2D = OCTAOJ2D + CH3

1.00E+14 7.94E+16 1.00E+14 2.00E+12 3.16E+16

0 0 0 0 0

0 80,280 0 0 83,070

[21] [39] Est. Est. [39]

H atom abstraction reaction examples MOCT + H = H2 + MOCT2J MOCT + O = OH + MOCT2J MOCT2D + CH3 = MOCT2D4J + CH4 MOCT2D + H = MOCT2D4J + H2

3.17E+04 2.20E+13 2.00E+11 6.55E+12

2.77 0 0 0

2276 3280 6800 4445

[27] [21] [21] [28]

Isomerization reaction example MOCT2J = MOCT5J

9.90E+07

1

17,300

[30]

Beta scission reaction examples C2H4 + MHEX6J = MOCT8J C7H15CO + CH2O = MOCTMJ MOCT2D4J = nC3H7 + MPEN24D C3H6 + MPEN2D5J = MOCT2D7J

2.00E+11 1.17E+01 1.00E+11 8.80E+03

0 0 0 2.48

7600 11,198 37,000 6130

[21] [27] [29] [35]

concerning the C–H bond breaking is supported by the work of El-Nahas et al. [38]. Their ab initio calculations showed the C–H bond strength varies at the different carbon positions of methyl butanoate (4, 3, 2, m) and as such the C–H bond breaking rate parameters from the terminal carbon, second carbon, and methoxy carbon positions for methyl octanoate were assumed to be the same as their analogous carbon positions for methyl butanoate regardless of the hydrocarbon chain length. As is frequently done with linear alkanes the central C–H bonds (position 7 through 3 of methyl octanoate) were assumed to be the same and have equivalent rate constants as the central C–H bond of methyl butanoate (carbon 3). For the C–C scission reactions within the alkyl side chain and between the alkyl side chain and the methyl ester group (bond 1 between carbon 1 and carbon 2) for methyl octanoate, the rate constants proposed by Dayma et al. [39] were adopted. For the C–O scission reaction of both fuels which yield either CH3 + acid based radical, or CH3O + a ketone radical, the reaction rate constants proposed by Dayma et al. [39] were also adopted. For the C–C scission reactions within the alkenyl side chain and between the alkenyl side chain and the methyl ester group (bond 1 between carbon 1 and carbon 2) for methyl trans-2-octenoate, rate constants have been estimated by group additivity methods. All C– C scission reactions, other than the scission that occurs at the b position from the double bond (between carbons 4 and 5 in Fig. 2) were first written as a recombination of radicals that was assumed to occur at the high pressure limit of 1E13. Next, the thermochemical data of each species was estimated, if not already available within the literature, using the group additivity methods of Benson [25] employed by THERGAS [26]. Finally, the reverse rate parameters were calculated using the assumed high pressure radical recombination forward rate constant and the calculated equilibrium rate constant from the thermochemical data. The calculation of the reverse rate parameters (the C–C scission) was achieved using CHEMRev [40]. 4.2. H atom abstractions by radicals and bimolecular initiations Hydrogen abstractions were written to occur via the following species: H, CH3, C2H3, C2H5, CH3O, CH3OH, O, OH, HO2, and O2. For all of the hydrogen abstraction reaction involving these radicals, except for H, the rate parameters purposed by Fisher et al. [21] for methyl butanoate were adopted based on the structure– reactivity relationships described earlier (for C–H bond breaking initiation reactions). For hydrogen abstractions via H atom for

2306

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

methyl octanoate from carbon numbers 8–3 the reaction rate constants were taken from Huynh and Violi [27] and assumed to be the same as their analogous (terminal or central) hydrogen abstractions via H atom on the alkyl chain of methyl butanoate. For hydrogen abstractions via H atom for methyl octanoate from carbon number 2 and the methoxyl carbon, the reaction rate constant was taken from Huynh and Violi [27] and assumed to be the same as the hydrogen abstractions via H atom from carbon number 2 and the methoxyl carbon of methyl butanoate respectively. The rational for this pattern of rate parameter adoption again was based on C–H dissociation energies calculated El-Nahas et al. [38] from which structure–reactivity relations were drawn as discussed in the previous reaction class (initiation reactions). For this class of reactions with methyl trans-2-octenoate the majority of the hydrogen abstraction rate parameters were again adopted or derived from (based on structure and reactivity relationships) Fisher et al. [21] with a few exceptions. H atom abstractions via H, OH and O from the fourth carbon of methyl octenoate were assumed to be the same as the H atom abstraction reactions via these same radicals from the third carbon of 1-hexene as proposed by Yahyaoui et al. [28] (see Fig. 4). The fourth carbon of methyl octenoate (seen in Fig. 5) is the adjacent carbon to the double bond and is analogous to the third carbon of 1-hexene (seen in Fig. 4). As is common practice with linear alkenes, the reaction rates involving hydrogen abstractions via various radicals from the carbon atom adjacent to the double bond were assumed to be equivalent regardless of the chain length but obviously dependant on the radical species involved in the reaction. Similarly, H atom abstractions via C2H3 from the fourth carbon of methyl octenoate were assumed to be the same as the H atom abstraction reaction from third carbon of 1-heptene by C2H3 as estimated from Westbrook et al. [29]. In addition, abstraction via CH3 from carbon positions 6–8 and 2 were estimated from Westbrook et al. [29] by the same rational discussed above.

which, yielding either alkenes or unsaturated methyl esters, the rate parameters were adopted from Fisher et al. [21] and assuming that all the b-scission reactions within the alkyl side chain have the same reaction rate. However, the b-scission reaction that occurs when the radical site is on the methoxy carbon was written as the recombination of formaldehyde and C7H15CO yielding MOCTMJ (reverse direction is the b-scission) and the rate constant was assumed to be the same as its analogous C4 methyl ester reaction from Huynh and Violi [27]. These two reactions are illustrated in Fig. 6 and the rates are assumed to be independent of the hydrocarbon chain length. Concerning b-scission reactions involving C–C cleavage in methyl octenoate radicals which, yield either alkenes or polyunsaturated methyl esters the rate constants were estimated from either Huynh and Violi [27], Westbrook et al. [29], or Herbinet et al. [35] based on structure and reactivity. The b-scission reaction that occurs when the radical site is on the methoxy carbon was written as the recombination of formaldehyde and C7H13CO yielding MOCT2DMJ (reverse direction is the b-scission) the rate constant was assumed to be the same as its analogous C4 methyl ester reaction from Huynh and Violi [27] (analogous to the reactions seen in Fig. 6 – second b-scission reaction in Table 1). The b-scission reaction that occurs when the radical site is on the fourth carbon (adjacent to the double bond) was assumed to have the same reaction rate constant as the b-scission that would occur from a 1-heptene fuel radical with the radical site being adjacent to the double bond. This reaction, illustrated in Fig. 7 (illustrating the third b-scission reaction in Table 1), was taken from Westbrook et al. [29]. b-scission reaction that occurs when the radical site is located between carbon numbers five and eight were adopted directly from their analogous methyl 2-decenoate reactions proposed by Herbinet et al. [35]. As is common practice with linear alkenes, the reaction rates involving b-scissions within the alkenyl side chain of methyl octenoate were assumed to be equivalent regardless of the hydrocarbon chain length.

4.3. Isomerization reactions The reaction rates for the isomerization of the seven different radical isomers of methyl octanoate were adopted directly from Togbe et al. [30]. Isomerization of any of the methyl octenoate fuel radical were assumed not to occur due to the strong preference of methyl octenoate to first decay through a C–C cleavage between carbon number four and five. 4.4. b-Scissions For b-scission reactions involving C–C cleavage in methyl octanoate radicals (written in reverse direction within the model)

2

6

4

1

3

5

Fig. 4. 1-Hexene with carbons numbers associated with the nomenclature from Fisher et al. [21].

m O

7 8

5 6

3 4

2

1 O

Fig. 5. Methyl trans-2-octenoate with carbons numbers associated with the nomenclature from Fisher et al. [21].

5. Modeling results, validation, and discussion There have been a number of recent experimental and kinetic modeling studies that have focused on the oxidation of biodiesel surrogates or surrogate components [21,22,30–36]. The majority of the experimental studies within the literature, which have been used to develop and validate the available kinetic models, have been performed at only slightly elevated reaction pressures (up to 10 atm) and with long reaction times of 70 ms–1.5 s. These studies have provided excellent data that have formed the foundation for this study but the current literature data are nevertheless incomplete. This study presented in both parts of this two parts series has extended the available data and provides a validated chemical kinetic model that encompasses practical compression ignition engine reaction conditions (P, T, s). The intermediate species profiles that were gathered experimentally from the HPST (Part 1 [5]) and JSR as well as the shock tube ignition delay data as a part of this study have been used to validate the developed chemical kinetic model over the wide range of conditions. 5.1. Shock tube ignition delay The saturated and unsaturated C7 hydrocarbons (heptane, heptene, and heptadiene) are analogs and representative of the hydrocarbon side chains of the saturated and unsaturated C8 methyl esters (methyl octanoate and methyl trans-2-octenoate) examined in this study. It is necessary to have a complete and clear understanding of the C7 hydrocarbon side chain chemistry of methyl octanoate and methyl trans-2-octenoate due to the fact

2307

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

CH2 O

C

H2C

O

O

O

CH2 O

C

H2C

O O

O

Fig. 6. Reaction rates for these two two b-scission reactions assumed to be equivalent and adopted from Huynh and Violi [27].

O O

CH2 CH

O

O

CH

CH2

Fig. 7. Reaction rates for these two b-scission reactions assumed to be equivalent and adopted from Westbrook et al. [29].

that the vast majority of the stable intermediate species formed from the C8 methyl esters under both the practical combustion conditions examined in the HPST (Part 1 [5]) as well as the JSR study present in this manuscript stem from the chemistry of the C7 hydrocarbon side chains. Thus, To further constrain and validate the C7 hydrocarbon model [20] which forms the foundation upon which the methyl ester sub-mechanisms were constructed, elevated pressure (10 atm), shock tube ignition delay studies of 1heptene and 1,6-heptadiene were conducted at stoichiometries U  0.5, 1, and 1.5 (experimental data provided as Supplementary data). The ignition delay validation by Chaos et al. [37] for the nheptane portion of the model remains valid due to the direct adoption of all C0 through n-heptane chemistry from Chaos et al. [37].

The modeling predictions and experimental data (which are included in the Supplementary material) can be seen in Fig. 8 through Fig. 13. For all stoichiometries and for both fuels, the model predicts quite well the experimental data with a small exception being a slight under-prediction at the highest temperature for both fuels at U  0.5. As expected, and similar to the ignition delay measurements of saturated and unsaturated fatty acid methyl esters [42], an increase in the ignition delay time is observed with increasing degree of unsaturation. Figure 14 clearly shows the trend of increasing ignition delay time with increasing degree of unsaturation. This trend is further supported by Caton et al. [43] and Murphy et al. [44] where these studies have shown experimentally that alkanes have shorter ignition delay times compared to alkenes (see Figs. 9–12).

1-Heptene Ignition Delay, =1.5

1-Heptene Ignition Delay, =1

1000

OH ( s)

OH ( s)

1000

100

10

100

10 5.8

6.0

6.2

6.4

6.6

6.8

104/T

5

7.0

7.2

7.4

7.6

(1/K)

Fig. 8. 1-Heptene ignition delay, P5 = 10 atm, U = 1.5, [C7H14] = 1248 ppm. Time taken to be 50% of peak OH formation. [j]-1-heptene, line represent model predictions.

6.2

6.4

6.6

6.8

7.0

7.2

7.4

7.6

7.8

104/T5 (1/K) Fig. 9. 1-Heptene ignition delay, P5 = 10 atm, U = 1, [C7H14] = 873.3 ppm. Time taken to be 50% of peak OH formation. [j]-1-heptene, line represent model predictions.

2308

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

1-Heptene Ignition Delay, Φ =0.5

1,6-Heptadiene Ignition Delay, =1

1000

τ OH (μ s)

τ OH (μ s)

1000

100

10

100

10 6.2 6.4 6.6 6.8 7.0 7.2 7.4 7.6 7.8 8.0 8.2 8.4

6.0

6.4

104/T5 (1/K) Fig. 10. 1-Heptene ignition delay, P5 = 10 atm, U = 0.5, [C7H14] = 447.6 ppm. Time taken to be 50% of peak OH formation. [j]-1-heptene, line represent model predictions.

7.2

7.6

8.0

Fig. 12. 1,6-Heptadiene ignition delay, P5 = 10 atm, U = 1, [C7H12] = 907.4 ppm. Time taken to be 50% of peak OH formation. [N]-1,6-heptadiene, line represent model predictions.

1,6-Heptadiene Ignition Delay,

1,6-Heptadiene Ignition Delay, =1.5

=0.5

1000

OH ( s)

1000

OH ( s)

6.8

104/T5 (1/K)

100

10

100

10 5.6

6.0

6.4

6.8

7.2

7.6

8.0

104/T5 (1/K) Fig. 11. 1,6-Heptadiene ignition delay, P5 = 10 atm, U = 1.5, [C7H12] = 1300 ppm. Time taken to be 50% of peak OH formation. [N]-1,6-heptadiene, line represent model predictions.

5.2. HPST results and validation The extension of the validation range of the C8 methyl ester model beyond the HPST data that were presented in Part 1 [5] of this series has not affected the HPST modeling prediction presented in Part 1. Therefore, only the extended validation will be show here. 5.3. JSR Four sets of experiments have been performed using the JSR setup with dilute mole fractions (1000–1500 ppm) of the fuels methyl octanoate and methyl trans-2-octenoate using nitrogen as the diluent. These four sets were performed at a nominal reaction pressure 10 atm, for a fixed reaction time of 0.7 s, while sweeping a temperature range of 550–1200 K at both fuel rich (U  2) and stoichiometric conditions. For all sets of experiments the carbon

6.0 6.2 6.4 6.6

6.8 7.0 7.2 7.4 7.6 7.8 8.0 8.2

104/T5 (1/K) Fig. 13. 1,6-Heptadiene ignition delay, P5 = 10 atm, U = 0.5, [C7H12] = 476.0 ppm. Time taken to be 50% of peak OH formation. [N]-1,6-heptadiene, line represent model predictions.

balances were good, better than 90% carbon recovered over the entire temperature range studied (see Supplementary Tables). Species were sampled from the JSR using a heated quartz probe and analyzed using both online and offline GC and GC/MS techniques. The variety of stable species formed in these experiments were sampled and quantitatively analyzed with the primary intermediates (by mole fraction) being: methane, ethane, ethane, acetylene, propene, butene, butadiene, carbon monoxide, carbon dioxide, formaldehyde, and butyraldehyde (butanal). No unsaturated fatty acid methyl esters were seen from the oxidation of the saturated fuel, methyl octanoate, and similarly no poly unsaturated methyl esters were seen from the oxidation of the mono unsaturated fuel, methyl trans-2-octenoate. The experimental profiles and modeling predictions for methyl octanoate, methyl trans-2-octenoate, and their intermediates can be seen in Fig. 15 through Fig. 23. Similar to the HPST results seen for the pyrolysis and oxidation of methyl octanoate and methyl trans-2-octenoate, increased amounts of acetylene (C2H2) are formed from the unsaturated fuel, methyl

2309

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

Methyl Octanoate, Φ =2

C7 Ignition Delay Comparison, 0.0015 0.0012

Mole Fraction

0.0009

OH (us)

1000

100

0.0006 0.0003 0.0000 0.010 0.008 0.006 0.004 0.002 0.000 700

10

800

104/T5 (1/K)

900

1000

1100

1200

Temperature (K)

6.0 6.2 6.4 6.6 6.8 7.0 7.2 7.4 7.6 7.8 8.0 8.2

Fig. 14. Comparison of C7 ignition delay times at P  10 atm, U = 1. [d]-heptane (Gauthier et al. [41]), [j]-1-heptene, [N]-1,6-hetpadiene.

Fig. 16. Methyl octanoate oxidation, P = 10 atm, U = 2, s = 700 ms, [methyl octanoate, U = 2] = 1580 ppm: [N]-methyl octanoate, []-CO, [h]-CO2, [x]-CH4, [d]C2H4, [+]-C2H6, [r]-CH2O. Lines represent model predictions [—]-methyl octanoate, [  ]-CO, [-]-CO2, [--]-CH2O, [—]-CH4, [- - -]-C2H4, [  ]-C2H6. Errorexp-data = ±5%.

trans-2-octenoate, for the JSR oxidation studies (see Fig. 15). Acetylene, as discussed in Part 1 [5] of this series, can be linked though a prompt NO formation mechanism such that an increase in acetylene formation has a corresponding effect on the increased NO formation seen during the combustion of biodiesel fuels in compression ignition engines. It is clear from Fig. 15 that a temperature window exists between approximately 850 K and 1050 K where the acetylene formation from the unsaturated fuel is greater than the acetylene formation from the saturated fuel. The maximum difference in the formation of acetylene between the two fuels is seen to occur at approximately 975 K and the mole fraction difference is on the order of a factor of 2 greater from the unsaturated fuel. These JSR experimental results support the experimental results from the HPST work (published in Part 1 [5] of this series) in that larger amounts of acetylene are formed from the unsaturated fuel compared to the saturated fuel. The complementary (to Fig. 14) HPST fuel decay and acetylene formation figure has been included as Supplementary material. In contrast to the HPST data in Part 1 [5], a significant shift in the reactivity of the two fuels is not seen

experimentally. This could be due to the high recirculation ratio that exists in the JSR system (a perfectly stirred reactor) but is not entirely clear. The lack of a shift in the reactivity between the unsaturated and saturated methyl esters coupled with difficulties in simulating some of the major stable intermediates seen in the JSR experiments (discussed in the following paragraphs) makes an analysis and comparison of the JSR acetylene formation pathways against the HPST acetylene formation pathways convoluted at best. Therefore, a presentation of the acetylene formation pathways for the JSR experiments as well as a comparison between it and the acetylene formation pathway analysis from the HPST is not discussed here. For the fuel rich oxidation (U  2) data seen in Fig. 16 through Fig. 19, the fuel decay is predicted within the experimental error. The modeling results for methyl octanoate decay, methane, formaldehyde, ethane, and to a lesser extent ethene formation are satisfactorily (Fig. 16a and b) reproduced while CO and CO2 are not so well predicted. For both CO and CO2, the developed model over-predicts the experimental mole fractions by a factor of two above 950 K at which point a large increase in the predicted mole

1.25

Fuel Decay and Acetylene Formation, JSR, =1 1.25 1.00

Normalized Mole Fraction

Normalized Mole Fraction

1.00

Fuel Decay and Acetylene Formation, JSR, =2

0.75 0.50 0.25 0.00 0.0150 0.0125 0.0100 0.0075 0.0050 0.0025 0.0000

0.75 0.50 0.25 0.00 0.10 0.08 0.06 0.04 0.02 0.00

600

700

800

900

1000

Temperature (K)

1100

1200

700

800

900

1000

1100

1200

Temperature (K)

Fig. 15. Fuel decay and acetylene formation comparison. [methyl octanoate, U = 2] = 1580 ppm, [methyl octanoate, U = 1] = 1250 ppm, [methyl octenoate, U = 2] = 1430 ppm, [methyl octenoate, U = 1] = 1380 ppm, s = 700 ms. [N]-methyl octanoate, [j]-methyl octenoate, [D]-C2H2 from methyl octanoate, [ ]-C2H2 from methyl octenoate. The lines are visual aids that represent curve fits to the experimental data.

2310

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

Methyl Octanoate,

=2

0.0005

0.00025

0.0004

0.00020

Mole Fraction

Mole Fraction

0.00010 0.00005 0.00000 0.00015 0.00012 0.00009 0.00006 0.00003 0.00000 700

800

900

1000

1100

1200

Fig. 17. Methyl octanoate oxidation, P = 10 atm, U = 2, s = 700 ms, [methyl octanoate, U = 2] = 1580 ppm: []-CH3CHO, [D]-C2H2, [}]-C3H6, [|]-1,3-C4H6, [.]-C4H8, [J]-C5H10, [.]-C6H12. Lines represent model predictions: [—]-CH3CHO, [—]-C3H6, [--]-C2H2, [-]-C4H8, [—]-1,3 C4H6, [-]-C5H10, [  ]-C6H12. Errorexp-data = ±5%.

fraction of each species is seen. Attempts to remove the jump in the predicted mole fractions of CO and CO2 at approximately 950 K by including new reactions or modifying the reaction rate constants of the existing reactions within the model did not resolve the discrepancy. For the remainder of the intermediate species seen (Fig. 16b) the qualitative trends are satisfactorily reproduced but there exists a quantitative over prediction of nominally a factor of two compared to the experimental results (Fig. 17a and b). The model performs similarly for the unsaturated fuel, methyl trans-2octenoate (Figs. 18 and 19). The modeling predictions for methyl trans-2-octenoate decay, methane, formaldehyde, ethane, and to a lesser extent ethene formation are satisfactorily (Fig. 18a and b) reproduced while CO and CO2 are not so well reproduced (similar to the predictions seen for methyl octanoate). For both species, the developed model over-predicts the experimental mole fraction by a factor of two above 950 K at which point a large jump in the predicted mole fraction of each species is seen. Again, attempts to remove the spike in the predicted mole fractions of CO and CO2 from methyl

Methyl Octenoate, Φ =2

a

0.0012

Mole Fraction

0.0009 0.0006 0.0003 0.0000 0.010

b

0.006 0.004 0.002 0.000 600

700

800

900

1000

1100

0.0002 0.0001 0.0000 0.00030 0.00025 0.00020 0.00015 0.00010 0.00005 0.00000

b

600

700

800

900

1000

1100

1200

1300

Temperature (K)

Temperature (K)

0.008

=2

0.0003

0.00015

0.0015

Methyl Octenoate,

a

1200

1300

Temperature (K) Fig. 18. Methyl octenoate oxidation, P = 10 atm, U = 2, s = 700 ms, [methyl octenoate, U = 2] = 1430 ppm: [j]-methyl octenoate, []-CO, [h]-CO2, [x]-CH4, [d]C2H4, [+]-C2H6, [r]-CH2O. Lines represent model predictions [—]-methyl octanoate, [  ]-CO, [-]-CO2, [--]-CH2O, [—]-CH4, [- - -]-C2H4, [  ]-C2H6. Errorexp-data = ±5%.

Fig. 19. Methyl octenoate oxidation, P = 10 atm, U = 2, s = 700 ms, [methyl octenoate, U = 2] = 1430 ppm: []-CH3CHO, [D]-C2H2, [}]-C3H6, [|]-1,3-C4H6, [.]-C4H8, [J]-C5H10, [.]-C6H12. Lines represent model predictions: [—]-CH3CHO, [—]-C3H6, [--]-C2H2, [-]-C4H8, [—]-1,3 C4H6, [-]-C5H10, [  ]-C6H12. Errorexp-data = ±5%.

trans-2-octenoate at approximately 950 K by including new reactions or modifying the reaction rate constants of the existing reactions within the model did not resolve the discrepancy. The remainder of the intermediate species for the fuel rich oxidation of methyl trans-2-octenoate are qualitative well reproduced but there exists a quantitative over-prediction for propene and 1,3butadiene (Fig. 19a and b). The modeling results for the stoichiometric oxidation of both fuels closely resembles the trends seen for U  2. It can be seen that, for the stoichiometric oxidation of these fuels (methyl octanoate and methyl trans-2-octenoate) which is shown in Fig. 20 through Fig. 23, the fuel decay is satisfactorily predicted by the model. The modeling results for methyl octanoate decay, methane, formaldehyde, ethane, and to a lesser extent ethene formation are satisfactorily (Fig. 20a and b) reproduced while, similar to the fuel rich oxidation results, CO and CO2 are not so well reproduced. Although, the CO trend has much improved compared to the fuel rich predictions (as well as the peak quantity prediction) the CO2 is over-predicted by a factor of two above 1000 K. Akin to the fuel rich results, the other intermediates are qualitatively predicted but large quantitative over-predictions of propene and 1,3butadiene exist. These same results are echoed in the prediction for methyl octenoate at U  1. Numerous efforts to reconcile the overpredictions of these species such as the inclusion of peroxy radical chemistry resulted in only minimal improvements. This model, and its predictions, represent the most comprehensive model, spanning large temperature, pressure, and reaction time regimes for both C8 saturated and unsaturated fatty acid methyl esters. Some of the modeling effort did result in increased accuracy for the JSR simulation but severely degraded the model’s ability to simulate the HPST experimental data presented in Part 1 [5] and since the goal of this work was to produce a single model to span both the HPST and JSR experimental data sets the current model represents the best possible model to span both sets. Due to the model’s inability to capture some intermediate species formation trends a sensitivity analysis or pathway analysis to examine the model’s predicted pathways or a species sensitivity to a specific reaction within the model would not yield any data from which conclusions could be drawn as to why the model does not capture some intermediate species trends. Nevertheless, the model does sufficiently simulate the decay of the saturated and unsaturated methyl esters fuels and thus allows for the model to be used in a predictive manner to generate reaction pathways for the fuels’ decay (Figs. 21 and 22).

2311

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

=1

Methyl Octenoate,

0.0020

=1

0.0016 0.0012 0.0008

Mole Fraction

Mole Fraction

Methyl Octanoate, 0.0018 0.0015 0.0012 0.0009 0.0006 0.0003 0.0000 0.010 0.008 0.006 0.004

0.0004 0.0000 0.006 0.004 0.002

0.002 0.000

0.000 700

800

900

1000

1100

700

1200

Fig. 20. Methyl octanoate oxidation, P = 10 atm, U = 1, s = 700 ms, [methyl octanoate, U = 1] = 1250 ppm: [N]-methyl octanoate, []-CO, [h]-CO2, [x]-CH4, [d]C2H4, [+]-C2H6, [r]-CH2O. Lines represent model predictions [—]-methyl octanoate, [  ]-CO, [-]-CO2, [--]-CH2O, [—]-CH4, [- - -]-C2H4, [  ]-C2H6. Errorexp-data = ±5%.

900

1000

1100

Fig. 22. Methyl octenoate oxidation, P = 10 atm, U = 1, s = 700 ms, [methyl octenoate, U = 1] = 1380 ppm: [j]-methyl octenoate, []-CO, [h]-CO2, [x]-CH4, [d]C2H4, [+]-C2H6, [r]-CH2O. Lines represent model predictions [—]-methyl octanoate, [  ]-CO, [-]-CO2, [--]-CH2O, [—]-CH4, [- - -]-C2H4, [  ]-C2H6. Errorexp-data = ±5%.

Methyl Octenoate, =1

Methyl Octanoate, =1

0.0004

0.00018 0.00015 0.00012 0.00009 0.00006 0.00003 0.00000

0.0003

Mole Fraction

Mole Fraction

800

Temperature (K)

Temperature (K)

0.00008 0.00006

0.0002 0.0001 0.0000 0.00010 0.00008 0.00006

0.00004

0.00004

0.00002

0.00002 0.00000

0.00000 700

800

900

1000

1100

1200

700

Temperature (K) Fig. 21. Methyl octanoate oxidation, P = 10 atm, U = 1, s = 700 ms, [methyl octanoate, U = 1] = 1250 ppm: []-CH3CHO, [D]-C2H2, [}]-C3H6, [|]-1,3-C4H6, [.]-C4H8, [J]-C5H10, [.]-C6H12. Lines represent model predictions: [—]-CH3CHO, [—]-C3H6, [-]-C2H2, [-]-C4H8, [—]-1,3 C4H6, [-]-C5H10, [  ]-C6H12. Errorexp-data = ±5%.

Figures 24 and 25 illustrate the predicted decay pathways for the two methyl ester fuels, methyl octanoate and methyl trans2-octenoate respectively. These pathway analyses were generated using a rate of production analysis (also known as a reaction flux analysis) at the reaction time (700 ms) and pressure (10 atm) of the JSR experiments. The reaction temperature of 770 K was chosen to examine the dominate reactions for the initial fuel decay. For the decay of the saturated methyl ester fuel, methyl octanoate, Fig. 24 shows that, similar to the HPST methyl octanoate decay pathways presented in Part 1 [5], that the majority of the fuel decays through hydrogen abstraction reactions (via various radical species) from the second carbon atom (see Fig. 2). This supports the work of El-Nahas et al. [38] where their ab initio calculations showed the weakest C–H bond within saturated fatty acid methyl esters to be at the second carbon position. In contrast to the similarities between the reaction pathways for the decay of methyl octanoate from both the JSR and HPST experiments, there exists a stark difference between the rich oxidation pathways for the decay of the unsaturated fuel, methyl trans-2-octenoate, from both the JSR and HPST.

800

900

1000

1100

Temperature (K) Fig. 23. Methyl octenoate oxidation, P = 10 atm, U = 1, s = 700 ms, [methyl octenoate, U = 1] = 1380 ppm: []-CH3CHO, [D]-C2H2, [}]-C3H6, [|]-1,3-C4H6, [.]-C4H8, [J]-C5H10, [.]-C6H12. Lines represent model predictions: [—]-CH3CHO, [—]-C3H6, [-]-C2H2, [-]-C4H8, [—]-1,3 C4H6, [-]-C5H10, [  ]-C6H12. Errorexp-data = ±5%.

The results from the HPST pathway analyses (Part 1 [5]) showed the dominate initial decay pathway for methyl trans-2-octenoate to be a C–C cleavage at the beta bond (as counted from the double bond), Fig. 25 illustrated the initial decay pathways to be hydrogen abstraction from the alkenyl side chain via O2. These fuel radical then further decay via beta-scission reactions yielding either an alkyl radical, alkene, methyl ester radical or a polyunsaturated methyl ester. This would seem to confirm the experimental JSR results where large amounts of alkanes and alkenes were measured although polyunsaturated methyl esters were not observed in these experiments. 6. Summary Stable species measurements from the fuel rich and stoichiometric oxidation of saturated and unsaturated C8 methyl esters, methyl octanoate and methyl trans-2-octenoate respectively have been experimentally obtained as a function of reaction pressure, temperature, and reaction time using a jet stirred reactor (JSR) at

2312

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313

MOCT2J 33.0%

MOCTMJ

MOCT3J 11.5%

5.1%

MOCT8J

MOCT

5.1%

11.3%

11.3%

MOCT4J

11.3%

11.3% MOCT5J

MOCT7J

MOCT6J Fig. 24. Methyl octanoate (MOCT) decay pathway analysis. All pathways represent the sum of the hydrogen abstraction routes via the radical: CH3, CH3O, H, HO2, O, and OH. T = 770 K, P = 10 atm, s = 700 ms.

MOCT2D7J

MOCT2D8J 10.5%

HPST (Part I [5]) and JSR (Part II – see Fig. 15). This increased acetylene formation seen during the JSR experiments can be correlated to increased prompt NO formation thereby contributing to the overall increased NOx formation that is generally seen from combustion systems which are fueled with biologically derived diesel fuel [1–4]. Second, although difficulties were encountered in modeling the stable intermediate species, specifically acetylene, from the JSR experiments, the experimental results from both the HPST [5] and the JSR (this study), which span a large experimental range (T, P, s), show the same trend of increased acetylene formation with increased degree of unsaturation, and thus implying increased prompt NOx. The combined HPST and JSR results are seen to thereby further support the hypothesis that the increased NOx formation from biodiesel fuels stems from chemical effects (chemical structure of the fuel) and not solely physical effects (adiabatic flame temperature, pressure, bulk modulus differences, ect.). Acknowledgments The authors gratefully acknowledge financial support for this research from the Combustion and Plasma Systems Division, National Science Foundation through award # CTS-0553439. The authors also gratefully acknowledge the NSF International Research which made this international collaboration possible. The authors thank Aurelie Barret, Andrea Comandini, Brad Culbertson, Alex Fridlyand and Soumya Gudiyella for their help with the experiments and modeling efforts. Appendix A. Supplementary material

26%

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2011.06.011.

MOCT2D References

12.3% 38% MOCT2D4J

12.3%

MOCT2D6J

MOCT2D5J Fig. 25. Methyl trans-2-octenoate (MOCT2D) decay pathway analysis. All pathways represent the hydrogen abstraction routes via O2. T = 770 K, P = 10 atm, s = 700 ms.

an elevated pressure of 10 atm. These experimental results have been used to extend the validation for the chemical kinetic model, presented in Part 1 [5]. In addition to the methyl ester JSR experimental data, results on the ignition delay of 1-heptene and 1,6heptadiene have been presented and used to constrain the foundation of developed model since these linear hydrocarbons represent analogs of the hydrocarbon side chains within methyl ester fuels. A comprehensive chemical kinetic model for the pyrolysis and oxidation of the two representative biodiesel components, methyl octanoate and methyl trans-2-octenoate, has been generated and now demonstrated to by predictive over a wide range of practical conditions. To put the JSR experimental and kinetic modeling results in perspective with respect to the prompt NO formation which was discussed in great detail in Part I [5] of this series, the following summary statements can be made. First, greater amounts of acetylene (C2H2) are formed from the unsaturated fuel, methyl trans-2-octenoate, compared to the saturated fuel, methyl octanoate, under both fuel rich and stoichiometric conditions in both the

[1] M. Graboski, R. McCormick, Prog. Energy Combust. Sci. (24) (1998) 125–164. [2] R. McCormick, M. Graboski, T. Alleman, A. Herring, K. Tyson, Environ. Sci. Technol. (35) (2001) 1742–1747. [3] M. Graboski, R. McCormick, T. Alleman, A. Herring, The Effects of Biodiesel Composition on Engine Emissions from a DDC Series 60 Diesel Engine, NREL Report #NREL/SR-510-31461, February 2003. [4] R. McCormick, C. Tennant, R. Hayes, S. Black, J. Ireland, T. McDaniel, A. Williams, M. Frailey, C. Sharp, Regulated Emissions from Biodiesel Tested in Heavy-Duty Engines Meeting 2004 Emission Standards, NREL Report # NREL/ CP-540-37508, May 2005, . [5] S. Garner, K. Brezinsky, Combust. Flame (2011). DOI: 10.1016/j.combustflame. 2011.04.014.. [6] P. Flynn, R. Durrett, G. Hunter, A. Zur Loye, O. Akinyemi, J. Dec, C. Westbrook, SAE Tech. Paper # 1999-01-0509, 1999. [7] A. Konnov, Combust. Flame (156) (2009) 2093–2105. [8] Y. Zhang, A. Boehman, Energy Fuels (21) (2007) 2003–2012. [9] C. Mueller, A. Boehman, G. Martin, SAE Tech Paper # 2009-01-1792, 2009. [10] R. Tranter, R. Sivaramakrishnan, N. Srinivasan, K. Brezinsky, Int. J. Chem. Kinet. (33) (2001) 722–731. [11] W. Tang, K. Brezinsky, Int. J. Chem. Kinet. (38) (2005) 75–97. [12] H. Mirels, Phys. Fluids (6) (1963) 1201–1214. [13] H. Mirels, AIAA J. (2) (1964) 84–93. [14] G. Dayma, S. Gail, P. Dagaut, Energy Fuels (22) (2008) 1469–1479. [15] P. Dagaut, M. Cathonnet, J. Rouan, R. Foulatier, A. Quilgars, J. Boettner, F. Gaillard, H. James, J. Phys. E (19) (1986) 207–209. [16] P. Dagaut, M. Cathonnet, Combust. Flame (113) (1998) 620–623. [17] A. El-Bakali, M. Braun-Unkhoff, P. Dagaut, P. Frank, M. Cathonnet, Proc. Combust. Inst. (28) (2000) 1631–1638. [18] P. Dagaut, A. Nicolle, Combust. Flame (140) (2005) 161–171. [19] C. Paillard, S. Youssefi, G. Dupre´, Prog. Astronaut. Aeronaut. (105) (1986) 394– 406. [20] S. Garner, R. Sivaramakrishnan, K. Brezinsky, Proc. Combust. Inst. (32) (2009) 461–467. [21] E. Fisher, W. Pitz, H. Curran, C. Westbrook, Proc. Combust. Inst. (28) (2000) 1579–1586. [22] S. Gail, S. Sarathy, M. Thomson, P. Dievart, P. Dagaut, Combust. Flame (155) (2008) 635–650. [23] R.J. Kee, F.M. Rupley, J.A. Miller, et al., CHEMKIN Collection, Release 3, sixth ed., Reaction Design Inc., San Diego, CA, 2000.

S. Garner et al. / Combustion and Flame 158 (2011) 2302–2313 [24] R.J. Kee, F.M. Rupley, J.A. Miller, M.E. Coltrin, J.F. Grcar, E. Meeks, H.K. Moffat, A.E. Lutz, G. Dixon-Lewis, M.D. Smooke, J. Warnatz, G.H. Evans, R.S. Larson, R.E. Mitchell, L.R. Petzold, W.C. Reynolds, M. Caracotsios, W.E. Stewart, P. Glarborg, C. Wang, C.L. McLellan, O. Adigun, W.G. Houf, C.P. Chou, S.F. Miller, P. Ho, P.D. Young, D.J. Young, D.W. Hodgson, M.V. Petrova, K.V. Puduppakkam, CHEMKIN Release 4.1, Reaction Design Inc., San Diego, CA, 2006. [25] S. Benson, Thermochemical Kinetics, second ed., Wiley: Interscience, NY, 1976. [26] C. Muller, V. Michael, G. Scacchi, G. Come, J. Chim. Phys. (92) (1995) 1154– 1178. [27] L. Huynh, A. Violi, J. Org. Chem. (73) (2008) 94–101. [28] M. Yahyaoui, N. Djebaili-Chaumeix, P. Dagaut, C. Paillard, S. Gail, Combust. Flame (147) (2006) 67–78. [29] C. Westbrook, W. Pitz, J. Warnatz, W. Jager (Eds.), Complex Chemical Reaction Systems: Mathematical Modeling and Simulation, Springer-Verlag, NY, 1987, pp. 39–54. [30] C. Togbe, G. Dayma, A. Mze-Ahmed, P. Dagaut, Energy Fuels (24) (2010) 3906– 3916. [31] P. Dagaut, S. Gail, M. Sahasrabudhe, Proc. Combust. Inst. (31) (2007) 2955– 2961. [32] W. Metcalfe, S. Dooley, H. Curran, J. Simmie, A. El-Nahas, M. Navarro, J. Phys. Chem. A (111) (2007) 4001–4014.

2313

[33] C. Togbe, J. May-Carle, G. Dayma, P. Dagaut, J. Phys. Chem. A (114) (2009) 3896–3908. [34] O. Herbinet, W. Pitz, C. Westbrook, Combust. Flame (154) (2008) 507–528. [35] O. Herbinet, W. Pitz, C. Westbrook, Combust. Flame (157) (2010) 893–908. [36] M. Hakka, P. Glaude, O. Herbinet, F. Battin-Leclerc, Combust. Flame (156) (2009) 2129–2144. [37] M. Chaos, A. Kazakov, Z. Zhao, F. Dryer, Int. J. Chem. Kinet. (39) (2007) 399– 414. [38] A. El-Nahas, M. Navarro, J. Simmie, J. Bozzelli, H. Curran, S. Dooley, W. Metcalfe, J. Phys. Chem. A (111) (2007) 3727–3739. [39] G. Dayma, C. Togbe, P. Dagaut, Energy Fuels (23) (2009) 4254–4268. [40] S. Rolland, J. Simmie, Int. J. Chem. Kinet. (37) (2005) 119–125. [41] B. Gauthier, D. Davidson, R. Hanson, Flame (139) (2004) 300–311. [42] C.S. Alexander, M. Ratcliff, National Renewable Energy Laboratory Report NREL/MP-540-43696, 2008. [43] P. Caton, L. Hamilton, J. Cowart, J. Eng. Gas Turbine Power (133) (2011). [44] M.J. Murphy, J.D. Taylor, R.L. McCormick, Compendium of Experimental Cetane Number Data, National Renewable Energy Laboratory, Report No. NREL/SR540-36805.