Detailed chemical kinetic mechanism for surrogates of alternative jet fuels

Combustion and Flame 158 (2011) 434–445

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

Detailed chemical kinetic mechanism for surrogates of alternative jet fuels Chitralkumar V. Naik a,⇑, Karthik V. Puduppakkam a, Abhijit Modak a, Ellen Meeks a, Yang L. Wang b, Qiyao Feng c, Theodore T. Tsotsis c a

Reaction Design, San Diego, CA 92121, USA Department of Aerospace and Mechanical Engineering, University of Southern California, Los Angeles, CA 90089, USA c Mork Family Department of Chemical Engineering and Material Science, University of Southern California, Los Angeles, CA 90089, USA b

a r t i c l e

i n f o

Article history: Received 4 March 2010 Received in revised form 27 July 2010 Accepted 22 September 2010 Available online 13 October 2010 Keywords: Fuel model Alternative jet fuels Fuel surrogate Reaction mechanism NOx emissions

a b s t r a c t Blends of n- and iso-alkane components are employed as surrogates for Fischer–Tropsch (F–T) and biomass-derived jet fuels. The composition of the blends has been determined based on data available for two F–T fuel samples obtained from different sources, using a systematic optimization approach. A detailed chemical kinetic mechanism for combustion of the surrogate blends has been assembled. The mechanism has been validated against fundamental experimental data. While drawing initially from other studies in the literature, the mechanism has been improved by enforcing self-consistency of the kinetic and thermodynamic data for the various surrogate-fuel components represented by the mechanism. These improvements have led to more accurate predictions of flame propagation, flame extinction, and NOx emissions. As part of the validation process, simulations were performed for a wide variety of experimental configurations, as well as for a wide range of temperatures and equivalence ratios for fuel/air mixtures. Comparison of the model predictions to the available literature data confirms the accuracy of the mechanism as well as of the approach for selecting the surrogate blends. Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction Fischer–Tropsch (F–T) fuels are of increased interest as alternative jet fuels. They can be appropriately tailored for specific applications. The feedstock for such fuels also varies from natural gas to bio-derived syngas. However, combustion and emissions characteristics of different F–T fuels are not yet understood in sufficient detail. It is critical to understand these characteristics when considering them as potential replacements of or blending components with conventional jet fuels, such as Jet-A. Accurate models will assist in understanding the behavior of these fuels and the effects of fuel variability during combustion. F–T fuels are less complex in composition than conventional fossil fuels that contain many different fuel classes and hundreds of components. Despite that, F–T fuels contain tens to hundreds of hydrocarbon components and their compositions vary depending on the different manufacturing processes. Accurate combustion and emissions modeling require good surrogate-fuels to represent the fuel chemistry of the practical fuel. This study focuses on gaining a better understanding of F–T fuels by developing accurate surrogate blends to represent the fuels and a detailed chemical kinetic mechanism for these surrogates. In developing the surrogate and its detailed mechanism, we have employed fundamental ⇑ Corresponding author. Fax: +1 858 550 1925. E-mail address: [email protected] (C.V. Naik).

experimental data to validate predictions of the combustion and emission characteristics. Measurements of NOx in laminar flames of various pure components and surrogates have been performed in this work. The rest of the fundamental data used for validation has been obtained from the literature. In this paper, we first discuss appropriate surrogates for specific fuels and then the development of a detailed reaction mechanism for these surrogates. Validation of model predictions of combustion behavior for the pure components and surrogate blends using various fundamental experimental data is then reported. This is followed by a discussion of important reaction pathways. In addition to F–T fuels, we have also considered applying the model to the biomass-derived jet fuels. Biomass-derived fuels are another form of alternative jet fuel. The composition of biomassderived jet fuel are expected to have similar composition [1,2] to that of the F–T fuels (discussed in the next section). Therefore, we expect the same models to be useful for both types of alternative jet fuels, although the surrogate blend composition may be different for different sources of fuels. Validation of the model against the limited fundamental experimental data available for a bio-jet fuel sample (R-8) has also been reported. 2. Identification of appropriate surrogate blends for F–T fuels Two different F–T fuels, namely Syntroleum S-8 and Shell GTL, have been studied in this work. Both were produced from natural

0010-2180/$ - see front matter Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.combustflame.2010.09.016

435

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

1.8 24.9 iso-alkane n-alkane cyclo-alkane

73.2 Fig. 1. Hydrocarbon class analysis of the S-8 F–T fuel in mol%.

60 S-8

50 Shell GTL

40

Mol %

gas using low temperature hydro cracking and isomerization process using cobalt [3]. Detailed composition of the fuels have been provided by the Air Force Research Laboratory [4]. Detailed analysis for S-8 composition is shown in Table 1 and the overall fuel class distribution in Fig. 1. Shell GTL fuel also includes all saturated longchain paraffins with up to two methyl side-chain components. The major distinctions between Shell GTL and S-8 fuels are that Shell GTL has a narrower carbon number distribution than that of S-8, as shown in Fig. 2, and the Shell GTL fuel contains higher amounts of n-alkanes (approximately 35 mol%) than those found in S-8 (approximately 26 mol%). In general, for both F–T fuel samples, the dominant components consisted of C7 to C16 n-and iso-paraffins with a narrower distribution in Shell GTL (C8 to C12). The F–T fuels contain a high concentration of iso-paraffins (65–74 mol%), which consist of only one or two methyl branches (–CH3) on a straight-chain alkane. In addition, based on the carbon number distribution shown in Fig. 2, more than half of these alkanes have a C10 to C12 chain. Capturing this range of carbon number components in a surrogate helps reproduce the overall variation of components in the fuels. The presence of lower carbon number components (C8) mainly affects the initial boiling point for the F–T fuel. Including components with C8, C10, and C12 saturated linear and slightly branched hydrocarbons in a surrogate provides a reasonable approximation of the overall chemical structure of F–T fuels. In order to generate a general-purpose surrogate to represent the physical and chemical properties, the focus of this study was

30 20 10 0 8

9

10

11

12

13

14

15

16

Carbon number Fig. 2. Carbon number distribution in the S-8 F–T fuel based on Ref. [3].

Table 1 Detailed analysis of fuel composition for S-8 based on the data provided by AFRL [4]. No.

Component name

mol%

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41

2-Methyl heptane 3-Methyl heptane 1,2,3-Trimethyl cyclopentane 2,5-Dimethyl heptane 4-Methyl octane 3-Methyl octane n-Nonane 3,5-Dimethyl octane 2,6-Dimethyl octane 4-Ethyl octane 4-Methyl nonane 2-Methyl nonane 3-Methyl nonane n-Decane 2-5-Dimethyl nonane 5-Ethyl-2-methyl octane 5-Methyl decane 4-Methyl decane 2-Methyl decane 3-Methyl decane n-Undecane x-Methyl undecane 3-Methyl undecane 5-Methyl undecane 4-Methyl undecane 2-Methyl undecane 2,3-Dimethyl undecane n-Dodecane 4-Methyl dodecane x-Methyl dodecane 2-Methyl dodecane x-Methyl dodecane n-Tridecane 4-Methyl tridecane 6-Propyl tridecane x-Methyl tridecane n-Tetradecane x-Methyl tetradecane 5-Methyl tetradecane n-Pentadecane x-Methyl tetradecane

0.62 0.84 1.85 2.17 4.80 2.53 3.11 1.98 1.45 1.98 3.65 1.95 2.65 3.93 2.25 1.94 2.52 2.17 2.93 3.03 4.64 3.05 2.20 3.25 2.00 2.05 2.32 4.97 1.78 1.43 2.48 2.45 3.33 1.60 2.02 2.04 2.99 2.30 1.38 1.98 1.39

to propose simple surrogates that not only represent combustion and emission characteristics of F–T fuels but also reflect some of the fuel composition and structural characteristics, as well as some of the physical properties like liquid density. Cetane Number is one important indicator of the combustion characteristics of jet fuels. Fundamental laminar flame properties that are governed by high-temperature chemical kinetics are also relevant for jet-engine combustion. NOx and soot emissions are additional factors that should be considered when designing the surrogate. Though physical properties are of lesser importance from a pure-combustion modeling point of view, it is desirable to capture also some physical properties. A single component surrogate may be able to describe one or a few combustion characteristics such as laminar flame speeds, but is not likely to be able to accurately capture a broader range of characteristics and certainly cannot reflect effects of differences in fuel composition like carbon number variation and boiling range. Capturing physical properties with surrogate is useful for Computational Fluid Dynamics (CFD) modeling where spray and evaporation are important along with fuel-combustion chemistry. In this study, we have limited the choice of components to three in order to limit the size of the detailed reaction mechanism and at the same time capture the range of carbon number variation, initial boiling point, and Cetane Number (CN) of F–T fuels. Another constraint in component selection for surrogates was the availability of sufficiently accurate detailed reaction mechanisms. We have selected the components for which there existed detailed reaction mechanisms in the literature that we used to assemble a unified master mechanism for the surrogates discussed in the next section. We selected a three-component blend of n-dodecane, n-decane, and iso-octane for both F–T fuels. The same blend is also tested for biomass-derived jet fuels as. n-Dodecane and n-decane are selected to represent the peaks in carbon number distribution as seen in Fig. 2. They also represent well the linear alkanes in practical fuels. While iso-decane and iso-dodecane could be good choices of components for representing the branched alkane class,

436

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

no detailed mechanisms were readily available. Developing such reaction mechanisms for such fuels can be quite time and resource consuming and out of scope of this work. Therefore, we chose isooctane as a representative of branched alkanes. Besides availability of a well studied detailed reaction mechanism, there are several benefits in selecting iso-octane as one of the surrogate components. First, the lower CN of iso-octane is necessary to capture the cetane rating of F–T fuels when mixed with other surrogate components n-dodecane and n-decane. Second, iso-octane is useful in capturing the initial boiling point of F–T fuels (near 403 K). Though the more highly branched iso-octane is not representative of the single-branched alkanes typically found in F–T fuels, we have found that lower concentrations of iso-octane can capture the effect of less-branched alkanes, which exist in high concentration in the real fuels, on overall combustion. Next, we have tuned or optimized (minimized differences between the blend properties and the targets) the composition of the three-component surrogate to match various target properties of two F–T fuel samples. For this purpose, we employed a software tool called the Surrogate Blend Optimizer (SBO) [5]. SBO determines an optimum surrogate blend composition, based on matching user-defined targets, which may include any combination of (1) chemical composition, (2) ignition quality (Cetane Number), (3) molar H/C ratio (relevant for determining the sooting propensity of the fuel), (4) lower heating value (LHV) (for overall heat release), (5) liquid fuel density at standard conditions, and (6) the True Boiling Point curve (ASTM D-2892 method). There are two methods implemented in SBO for optimization: global optimization based on a Genetic algorithm, followed by local optimization based on a Direction Set algorithm. Various estimation methods for Cetane Number as well as calculation methods for other properties have been implemented in SBO. True Boiling Point (TBP) curves are slightly different from distillation curves (ASTM D-86 method). However, both methods yield very close T50 (boiling point for 50% evaporation) [6]. Though SBO does not reproduce phase equilibrium, as suggested by Slavinskaya et al. [7], we expect that matching all the six targets including several points of the boiling point curve should provide a reasonable representation of both the chemical and physical qualities of conventional and alternative fuels. To generate optimal 3-component blends for the Shell GTL and the S-8 fuels, we used certain measured properties as targets, which are listed in Table 2. The SBO generated optimal surrogate compositions for the two F–T fuels. Note that these compositions are not unique, and that other blends may exist that can match the specified targets. The surrogate blend composition and associated properties are shown in Table 3 [8]. These surrogate blends are designed by the SBO to mimic not only the combustion behavior of the F–T fuels, but also the emissions behavior and. In the validation section, the surrogates have been validated for several combustion properties that are useful to validate the applicability for jet-engine simulations including NOx emissions. Though no explicit validation of sooting behavior is performed in this work, representing the molar H/C ratio and fuel-structure characteristics makes it likely that the surrogate will mimic the sooting behavior of F–T fuels.

Table 2 Measured properties used as targets for the two F–T fuels [4]. Target criteria

Shell GTL

S-8

Cetane Number H/C molar ratio Lower heating value (MJ/kg) ASTM D-86 T50 distillation point (K) Density (g/cm3)

61 2.17 44.2 445 0.736

60 2.17 44.1 474 0.755

Table 3 Composition of the three-component surrogate optimized by Surrogate Blend Optimizer for the two F–T fuel samples. Property

Shell GTL

S-8

iso-Octane (mol%) n-Decane (mol%) n-Dodecane (mol%) Cetane Number H/C molar ratio Lower heating value (MJ/kg) ASTM D-86 T50 distillation point (K) Density (g/cm3)

28 61 11 61 2.21 44.55 404 0.723

32 25 43 61 2.20 44.45 447 0.729

In addition to two F–T fuel samples, one sample of bio-jet fuel, R-8 was also considered in this study. Since detailed fuel composition of the bio-jet fuel, R-8 was not available; it was not possible to derive an optimized surrogate blend for it. Based on Refs. [1,2] that indicated that the R-8 composition was similar to F–T fuels, we used the same surrogate as that for S-8 for modeling the R-8 data. 3. Development of a detailed reaction mechanism for the F–T fuel surrogates For jet-engine combustion, usually only high-temperature (>1000 K) kinetics are relevant. Therefore, low-temperature chemistry can be omitted from the master mechanism, which helps reduce the size of the mechanism. A detailed chemical kinetic mechanism for high-temperature combustion of the surrogate blends has been assembled starting from component mechanisms reported in the literature. We assembled the master mechanism in stepwise fashion: First, assemble a single mechanism containing all surrogate components using the individual component mechanisms from the literature. Second, add sub-mechanisms for NOx and Polycyclic Aromatic Hydrocarbons (PAH) for soot predictions. Third, establish self-consistency in the mechanism and then improve the master mechanism to achieve more accurate predictions. These steps are described below. 3.1. Detailed reaction mechanisms for the combustion of surrogate components For the n-alkanes, we started with the recent mechanism of Westbrook et al. [9] that was designed for combustion of all linear alkanes from n-heptane up to n-hexadecane. The n-dodecane version of this mechanism used in this work contained 1283 species and 5030 reactions. The mechanism contains 25 reaction classes that describe fuel oxidation in low- and high-temperature regimes [9]. As mentioned previously, application to jet-engine combustion only requires high-temperature kinetics above 1000 K. Reactions that only apply to temperatures below 1000 K can be ignored without loss of accuracy; removing them provides a smaller size of the mechanism. We therefore removed all reactions pertaining to the low-temperature kinetics, specifically reaction classes 10–25 as defined by Westbrook et al. [9]. However, at high-pressure operating conditions of jet engines, the low-temperature kinetic regime may extend slightly towards higher temperatures [10]. To maintain the accuracy at high pressures, then, we kept all the reactions for both low- and high-temperatures involving all C0 (hydrogen) to C2 hydrocarbons. We extracted the high-temperature sub-mechanism for iso-octane using the same reaction-class-based approach, from the mechanism of Curran et al. [11]. This sub-mechanism was then merged with the high-temperature n-alkanes mechanism. The resulting mechanism contains all the necessary reactions for studying the high-temperature oxidation of all three surrogate-fuel components identified for F–T fuels. The mechanism still maintains

437

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

all reaction classes defined by Westbrook et al. and Curran et al. and contains 564 species and 3556 reactions. Note that the mechanism is applicable for not only the three F–T fuel components chosen in the previous section, but also for all linear alkanes from n-heptane to n-dodecane. There are two important points to note about the assembled master mechanism: The mechanism is self-consistent and the reaction rates are based on consistent reaction-rate rules. These facts increase the reliability and predictive capability of the mechanism. The Westbrook mechanism [9] did not contain the transport parameters required for flame simulations, such as the Lennard– Jones parameters required for CHEMKIN calculation of transport properties [12]. These transport parameters were obtained either from literature sources where available or estimated. CHEMKINcompatible transport parameters for alkanes up to C8 were obtained from several publications by Curran et al. [11,13]. Those for remaining alkanes from C9 to C12 in the surrogate mechanism were calculated based on critical-properties estimation using Joback’s group contribution method [14]. 3.2. NOx sub-mechanism To predict NOx formation and destruction during combustion, a comprehensive NOx sub-mechanism was first assembled based on multiple literature sources. The NOx sub-mechanism in GRI-mech 3.0 [15] was well tested for flame conditions but not for mid- to lower-temperatures or for high-pressure conditions. Recently several researchers developed detailed NOx mechanism that are applicable for various specific conditions, such as mid-temperature range, where HCN chemistry is important [16] and conditions at lowertemperatures and high pressures where NOx interacts with hydrocarbons [17]. To form a comprehensive mechanism that is applicable to not only high-temperatures but also to high-pressure conditions in jet engines, we merged three different mechanisms that have been developed and tested for different specific conditions. These mechanisms are the NOx subsets from GRI-mech 3.0, HCN chemistry from Dagaut et al. [16], and NOx-hydrocarbon sensitization chemistry from Rasmussen et al. [17]. These three different subsets of reactions cover the pathways applicable for high-temperatures, mid-temperatures (1200–1500 K), and high-pressure ranges, respectively. The resulting comprehensive NOx sub-mechanism was then added to the combustion mechanism for F–T surrogates assembled in the previous step. The comprehensive NOx submechanism adds 33 species and 298 reactions that were not included in the surrogate combustion mechanism, which increased the size of the master mechanism to 597 species and 3854 reactions.

and emissions chemistry. Addition the PAH sub-mechanism increased the size of the master mechanism by 156 species and 814 reactions, for a total of 753 species and 4668 reactions in the master mechanism. 3.4. Improvement of the master mechanism The initial surrogate mechanism significantly overpredicted the laminar flame speeds and the extinction strain rates measured for individual surrogate components. We also tested the mechanism for a few smaller hydrocarbons such as n-heptane where more experimental data are available. Since the mechanism is self-consistent and based on rate-rules, consistent trends are expected among all large components of similar hydrocarbon class. For example, Fig. 3 illustrates the laminar flame speeds results for n-heptane, using the initial master mechanism described in the previous steps. Similarly, Fig. 14 (discussed later) illustrates the overprediction by the initial master mechanism for the extinction strain rates for n-dodecane. Since accurate predictions of fundamental laminar flame experiments are critical for application under jet engine conditions, we performed rate and sensitivity analyses to identify the areas where the mechanism needed improvements. Based on the analyses, it was found that the core mechanism (C0 to C2 hydrocarbons) had great impact on predictions of these flame properties. Ji et al. [21] also concluded the same. We also performed extensive validation of hydrogen, CO, methane, and ethane combustion in a separate study [22] and found some discrepancies in these predictions as well. Therefore, in order to improve the predictions for laminar flame properties for F–T surrogate n-alkanes, we improved the core mechanism. This resulted in several key updates to the mechanisms. First, explicit reverse-rate specifications inherited from the parent source mechanism [9,11] used in the core mechanisms were removed. This allowed enforcement of thermodynamic consistency between forward-rate and reverse-rate calculations to ensure microscopic reversibility. Second, rate constants for several key reactions as shown in Table 4 were updated. These updates were based on results of other studies and were within the uncertainty range for the specific reactions considering experimental uncertainties as well as scatter among several sets of experimental data [23]. Though individual reactions have not been discussed here, their sources are listed in Table 4. In addition, transport parameters for some of the smaller hydrocarbons in the core mechanism were updated based on work by Wang and co-workers [24], although this did not have a significant impact on predictions.

3.3. PAH sub-mechanism Laminar flame speed (cm/s)

It is well known in combustion literature that Polycyclic Aromatic Hydrocarbons (PAH) leads to soot formation. To get an insight into sooting tendencies of F–T fuels, we assembled a sub-mechanism describing PAH formation and growth chemistry. Several literature mechanisms exist that describe PAH chemistry [18–20]. One popular mechanism is by Appel et al. [18] that was developed to predict soot formation in ethylene flames. It includes formation of the first aromatic ring (benzene) and further growth based on the HACA (Hydrogen Abstraction C2H2 Addition) scheme. Skjoth-Rasmussen et al. [19] developed a mechanism to describe PAH formation in methane flames. Recently Zhang et al. [20] included several other possible pathways leading to benzene formation from smaller hydrocarbon fuels. We have merged these three mechanisms to form a comprehensive PAH formation and growth sub-mechanism. This PAH sub-mechanism was then added to the master mechanism (with NOx) developed in the previous steps, creating a comprehensive, self-consistent master mechanism containing both combustion

55 50

n-Heptane/air at 298K and 1 atm

45 40 35 30 25 20

Davis and Law, 1998 Initial source mechanism Current mechanism

15 10 0.6

0.8

1.0

1.2

1.4

1.6

Equivalence ratio Fig. 3. Comparison of laminar flame speeds of n-heptane/air measured at 298 K and 1 atm by Davis and Law [32] (symbols) and those predicted using the initial source mechanism (dashed line) and the current improved mechanism.

438

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

Table 4 Updated rate coefficients for key reactions in the core mechanism. Rate coefficients are in units of cal, mol, cm3, K units. Rate constants are calculated as k = ATnExp( Ea/RT) where T represents temperature and R represents gas constant. No.

Reaction

A

n

Ea

Reference

1 2

H + O2 = O + OH CO + OH = CO2 + H Duplicate

1.97E+14 7.05E+04 5.76E+12

0 2.053 0.66

16,540 356 332

[42] [43]

3

HCO + M = CO + H + M

1.87E+17

1

17,000

2  k [41], adjusted collision efficiencies

4

H + O2(+M) = HO2(+M) Low Troe: 0.5 1e 30 1e + 30 Collision efficiencies: O2/0.85/H2O/11.89/CO/1.09/CO2/2.18/AR/0.40/

5.12E+12 6.33E+19

0.44 1.4

0 0

[44], M = n2  1.1

5

H + OH + M = H2O + M Collision efficiencies: H2/0.73/H2O/12.0/CO/1.90/CO2/3.8/AR/0.38/

4.40E+22

2

0

2  k [15]

6

CH3 + H(+M) = CH4(+M) Low Troe: 0.7830 74.00 2941.00 6964 Collision efficiencies: H2/2.0/H2O/6.0/CH4/2.0/CO/1.5/CO2/2.0/C2H6/3.0/AR/0.7/ CH2(S) + O2 => H + OH + CO CH2(S)+O2 = CO + H2O CO + HO2 = CO2 + OH OH + HO2 = H2O + O2 Duplicate

1.27E+16 2.48E+33

0.63 4.76

383 2440

[15]

2.80E+13 1.20E+13 1.57E+05 6.67E+28 2.51E+12

0 0 2.18 4.73 2

0 0 17,943 5503 40,000

[15] [15] [45] [45], re-fitted

0 10,840 0 0 0 0 1580 6685

[41] [15] This work This work [42] This work [15]

Collision efficiencies: H2/2.0/H2O/12.0/CO/1.75/CO2/3.6/

7 8 9 10 11 12 13 14 15 16 17

HCO + H = CO + H2 CH4 + H = CH3 + H2 C2H5 + H = 2CH3 C2H5 + H = C2H4 + H2 C2H3 + H = C2H2 + H2 C2H3 + H = H2CC + H2 C2H5 + H(+M) = C2H6(+M) Low Collision efficiencies: H2/2.0/H2O/6.0/CH4/2.0/CO/1.5/CO2/2.0/C2H6/3.0/AR/0.7/

1.20E+14 6.60E+08 9.00E+13 2.00E+12 1.20E+13 8.00E+12 5.21E+17 1.99E+41

0 1.62 0 0 0 0 0.99 7.08

18

CH3 + CH3(+M) = C2H6(+M) Low Troe: 0.5325 151.0 1038.00 4970.0 Collision efficiencies: H2/2.0/H2O/6.0/CH4/2.0/CO/1.5/CO2/2.0/C2H6/3.0/AR/0.7/

2.12E+16 1.77E+50

0.97 9.67

620 6220

[15]

19

C2H4 + H(+M) = C2H5(+M) Low Troe: 0.569 299.0 9147.0–152.40 Collision efficiencies: H2/2.0/H2O/6.0/CH4/2.0/CO/1.5/CO2/2.0/C2H6/2.0/AR/0.7/

1.37E+09 2.03E+39

1.463 6.64

1355 5769

[46], adjusted collision efficiency

20

C3H5 A + H(+M) = C3H6(+M) Low Troe: 0.020 1096.6 1096.6 6859.5 Collision efficiencies: H2/2/H2O/6/CH4/2/CO/1.5/CO2/2/C2H6/3/AR/0.7/

2.00E+14 1.33E+60

0

0 5968

[47]

12

C2H3 + CH3(+M) = C3H6(+M) Low Troe: 0.175 1340.6 60000.0 10139.8 Collision efficiencies: H2/2/H2O/6/CH4/2/CO/1.5/CO2/2/C2H6/3/AR/0.7/C2H2/3.00/C2H4/ 3.00/

2.50E+13 4.27E+58

0

0 9770

[48]

11.9

21

The improved detailed reaction mechanism for F–T surrogates, which contains 753 species 4668 reactions in CHEMKIN-compatible format, is provided in Supplementary material. The mechanism is also applicable for combustion of other fuel components, such as n-pentane, n-hexane, iso-hexane, n-heptane, n-octane, n-nonane, n-undecane, etc., which are not discussed explicitly here. In addition, a smaller version of the improved F–T surrogate mechanism excluding the PAH chemistry (Section 3.3), containing 597 species and 3854 reactions, is also provided in Supplementary material. 4. Validation The three-component F–T surrogate mechanism has been validated using several fundamental experimental data available in the literature. In addition, data on NOx emissions from laminar flames were obtained at the USC Flame Facility, using a counterflow burner assembly. The experimental procedures have been reported elsewhere [25] and a sketch is shown in Fig. 4. The

comparisons of predictions and measurements are organized based on the types of fundamental experiments considered and summarized in Table 5. The goal of the validation was both to verify the behavior of the individual components as well as to test the F–T fuel-surrogate definition against experiments that used actual F–T fuel samples as well as the bio-jet sample. The complete master mechanism has been used for simulations presented here, except for extinction strain rate simulation, as discussed later. In addition to the results reported here, the mechanism was also validated for the smaller core components such as H2, H2/CO, methane, and ethane. These results have been published elsewhere [22], and some of the validation of the core mechanism has been included in Supplementary material. 4.1. Autoignition delay times Ignition-delay times for iso-octane have been measured behind reflected shocks by Davidson et al. [26] and by Vermeer et al. [27].

439

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445 1.E-02 iso-Octane/O2/Ar 1.5 to 2 atm

Ignition time (s)

1.E-03

1.E-04 Davidson et al., phi 0.5, ~93% dilution Davidson et al., phi 1, ~93% dilution Davidson et al., phi 2, ~93% dilution Model, phi 0.5, ~93% dilution Model, phi 1,~93% dilution Model, phi 2,~93% dilution Vermeer et al., phi 1, 70% dilution Model, phi 1, 70% dilution

1.E-05

1.E-06 0.55

0.6

0.65

0.7

0.75

0.8

0.85

1000/T (1/K)

Fig. 4. Sketch of the atmospheric pressure counterflow burner assembly at the USC Flame Facility used for NOx measurements from laminar flames. Burners are separated by 14 mm. Heated vaporized fuel/air mixture at 403 K is injected from the bottom burner and room temperature nitrogen is injected from the top burner.

In this work, we modeled these experiments using the CHEMKINPRO [12], transient, zero-dimensional closed homogeneous reactor model under constant volume conditions. Figure 5 compares the calculated ignition-delay times with the experimental data of Davidson et al. [26]. The figure shows the effects of varying the equivalence ratio on the ignition-delay times, over a temperature range of 1300–1750 K. The model predictions agree well with the experimental data. They also show that the fuel-lean mixtures ignite earlier than the stoichiometric and fuel-rich mixtures at these temperatures. Figure 5 also compares the model predictions with the experimental data of Vermeer et al. [27]. The experimental data are taken in the pressure range of 1.7–2.5 atm and the stoichiometric iso-octane/O2 mixture is diluted with 70% Ar. The model predictions have been performed at 2 atm, and the results agree well with the experimental data over the temperature range shown. Figure 6 compares the calculated ignition-delay times with the experimental data of Pfahl et al. [28] at various equivalence ratios, and at a pressure of 13 atm. The figure also includes the data of Zhukov et al. [29] for a stoichiometric n-decane/air mixture at the same pressure. The predictions agree well with the experimental data over the temperature range shown. The model also captures the minor shifts in trends at different equivalence ratios. Figure 7 compares the calculated ignition-delay times with the experimental data of Vasu et al. [30] for (1) n-dodecane/air mixture at a pressure of 20 atm for equivalence ratios of 0.5 and 1.0, and (2) 1000 ppm of n-dodecane with O2 and Ar at 16 atm and an equivalence ratio of 0.5. The ignition times for n-dodecane/O2/Ar mixtures are predicted well. The predictions for fuel/air mixtures at temperatures above 1100 K are within a factor of two. However, the mechanism could not capture the faster ignition for fuel/air measured near 1000 K. A lack of low-temperature kinetics involv-

Fig. 5. Effect of equivalence ratio and dilution on iso-octane ignition-delay times. Comparison of calculated values with the experimental data of Davidson et al. [26] at 1.5 atm at 93% dilution, and those of Vermeer et al. [27] at 2 atm at 70% dilution.

ing hydroperoxides of fuel species in the current mechanism is one of the main reasons for this deviation at low-temperatures. It should be noted, however, that for applications to jet engines, high-temperature kinetics dominates and any discrepancy at low-temperatures does not reduce the accuracy. In addition, the measured ignition times for fuel/air mixtures can also be affected by the pressure rise in the shock-tube during the induction time which will cause the reaction to speed-up. To capture such experimental effects, the model needs to be further modified, but this is beyond the scope of the present paper. Another possibility is due to erroneous experimental measurements with n-dodecane/air mixture. The sub-mechanism of n-dodecane follows the same raterules as that of n-decane. It is unlikely for the predictions to be inaccurate for n-dodecane while those for n-decane are good as shown in Fig. 6. 4.2. Autoignition temperatures Bieleveld et al. [31] measured the autoignition temperature of various fuels in a counterflow burner assembly. In their experiments, they increased the temperature of the air until the fuel auto-ignited. The experimental data on autoignition temperature are only available for one of the surrogate components: iso-octane, as a function of strain rates for a fixed fuel mass fraction of 0.4 as shown in Fig. 8. However, they also measured the data for n-heptane. Since n-heptane is indirectly test of the quality of the mechanism built on rate-rules, we also used n-heptane data for extending the validation of the mechanism. Simulations of these experiments have been performed using the CHEMKIN-PRO Opposed-flow Flame model [12], by gradually increasing the temperature of air. Comparisons of the predicted and measured autoignition temperatures are shown in Fig. 8. As shown in the figure, the model captures the autoignition temperatures of n-heptane and iso-octane very well.

Table 5 Fundamental experimental data sets used for the validation of the surrogate mechanism. Measured property

iso-Octane

n-Decane

n-Dodecane

F–T fuels (S-8 and Shell GTL)

Bio-jet fuel (R-8)

Autoignition delays Autoignition temperatures Laminar flame speeds Species concentrations in burner stabilized flames Premixed flame extinction strain rates NOx emissions

[27,49] [31] [32–34] – – –

[28,29] – [21,35] [50] – This work

[30] – [21,35] – [21] This work

– – [36] – [36] This work

– – [36] – [36] This work

440

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

are closer to those of iso-octane at lower strain rates, and shift towards those of n-heptane at the higher strain rates.

1.E-02

4.3. Laminar flame speeds 1.E-03

Pfahl et al., phi 0.5 Model, phi 0.5 Pfahl et al., phi 1 Model, phi 1 Pfahl et al., phi 2 Model, phi 2 Zhukov et al., phi 1

1.E-04

1.E-05 0.7

0.75

0.8

0.85

0.9

0.95

1

1000/T (1/K) Fig. 6. Calculated n-decane/air ignition-delay times at 13 atm compared with the data of Pfahl et al. [28]. The data of Zhukov et al. [29] for stoichiometric conditions (filled triangles) are also shown on the plot.

1.E-02 n-Dodecane

Ignition time (s)

1.E-03

1.E-04 Model, fuel/air, 20 atm, phi 0.5 Model, fuel/O2/Ar, phi 0.5, 16 atm Model, fuel/air, 20 atm, phi 1 Vasu et al., fuel/air, 20 atm, phi 0.5 Vasu et al., fuel/air, 20 atm, phi 1 Vasu et al., fuel/O2/Ar, phi 0.5, 16 atm

1.E-05

1.E-06 0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

1000/T (1/K) Fig. 7. Calculated ignition-delay times for n-dodecane/air at 20 atm and n-dodecane/O2/Ar at 16 atm compared with the data of Vasu et al. [30].

Autoignition temperature (K)

1300

Fuel mass fraction 0.4 Fuel nozzle temperature 468 K, 1 atm

Experimental data of iso-octane/air laminar flame speeds have been collected from Davis and Law [32], Huang and Sung [33], and Kumar et al. [34]. The model predictions are compared with these data in Fig. 9. The uncertainty in experimental velocity measurements has been reported to be 1–2 cm/s by Davis and Law [32]. The unstretched laminar flame speeds have been determined using linear or non-linear extrapolation to zero stretch, and the differences between the two extrapolation techniques has been reported to be at least 1–2 cm/s as seen in the data of Davis and Law [32] as well as that of Huang and Sung [33]. Linear extrapolation has been used by Davis and Law [32], Kumar et al. [34] and Huang and Sung [33], and non-linear extrapolation techniques has also been used by Davis and Law [32]. Comparing the simulations with the data from various researchers, as shown in Fig. 9, the model predicts the laminar flame speeds well for the entire range of equivalence ratios. However, the agreement on the fuel-lean and stoichiometric conditions are better with the data of Huang et al. whereas on the fuel-rich side they agree better with the nonlinear-extrapolated data of Davis et al. Considering the uncertainties in the experimental data, the predictions are reasonably good. Accurate predictions for other n-alkanes, as discussed below, support this conclusion. Comparisons of the predicted laminar flame speeds of n-decane and n-dodecane with the experimental data of Ji et al. [21], as well as the data of Kumar et al. [35], are shown in Figs. 10 and 11 respectively. The peak flame speeds occurs at equivalence ratio of 1.1. The mechanism shows excellent agreement over a broad range of equivalence ratios in both cases with the data of Ji et al. The data of Kumar et al. on fuel-rich conditions are consistently higher than the other data and predictions. This discrepancy is likely due to uncertainties in the extrapolation technique used by Kumar et al. The non-linear extrapolation technique used by Ji et al. results in more accurate derivation of laminar flame speeds at zero stretch. A detailed discussion by Ji et al. [21] explains various uncertainties in the experimental data arising from different extrapolation techniques and proves the higher level of fidelity in their data. Next, the surrogate blends defined in Table 3 were used to test the model against data measured by Ji et al. [36] for both the F–T

1250

40 35

1200 Bieleveld et al., n-heptane Bieleveld et al., iso-octane Model, n-heptane Model, iso-octane Model, F-T (S-8) surrogate

1150

1100 200

300

400

500

600

700

Strain rate (1/s)

Laminar flame speed (cm/s)

Ignition time (s)

n-Decane/air, 13 atm

30 25 20 15 Davis et al., non-linear extrapolation Davis et al., linear extrapolation Huang et al. Kumar et al. Model

10 5

Fig. 8. Comparison of predicted and measured autoignition temperatures for fuel/air mixtures of n-heptane and iso-octane. Symbols represent the experimental data from Bieleveld et al. [31], and the lines represent the predictions. Predictions for F–T fuel using S-8 surrogate are shown for comparison.

To predict the autoignition behavior of F–T fuels, simulations were performed using the three-component surrogate blend for the S-8 fuel. As shown in Fig. 8, autoignition temperatures for F–T fuels

0 0.6

0.8

1

1.2

1.4

1.6

Equivalence ratio Fig. 9. Laminar flame speeds of iso-octane/air at 298 K and 1 atm. Experimental data from various studies [32–34] are compared with the model predictions over a range of equivalence ratios. The data from Davis and Law [32] also demonstrate the effect of linear and non-linear extrapolation methods used to derived the data from measurements.

441

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445 70

80

Fuel/air

70

Laminar flame speed (cm/s)

Laminar flame speed (cm/s)

n-Decane/air 403K, 1 atm

60 50 40 30

Ji et al. Kumar and Sung

20

403 K, 1 atm

60 50

40 Model, S-8

30

Model, Shell GTL Data, S-8

20

Data, Shell GTL

Model

10 0.6

0.8

1

Data, R-8

1.2

1.4

10 0.04

1.6

0.05

0.06

Equivalence ratio Fig. 10. Comparison of predicted laminar flame speed of n-decane/air at 403 K and 1 atm with those measured by Ji et al. [21] and Kumar et al. [34].

0.1

0.11

Daute et al., n-decane Model, n-decane Daute et al., O2 Model, O2 Daute et al., CO Model, CO Daute et al., CO2 Model, CO2 Daute et al., C2H4 Model, C2H4 Daute et al., C2H2 Model, C2H2 Daute et al., H2 Model, H2

0.18 0.16

50

0.14

40 30

Kumar and Sung Ji et al., Linear extrapolation Ji et al., Nonlinear extrapolation

20

0.12 0.1 0.08 0.06 0.04

Model

10 0.6

0.09

0.2 n-Dodecane/air 403 K, 1 atm

Mole fraction

Laminar flame speed (cm/s)

0.08

Fig. 12. Comparison of predicted flame speeds for the two F–T fuel surrogates at 403 K and 1 atm to the experimental data measured by Ji et al. [36].

70 60

0.07

Fuel/air mass ratio

0.02 0.8

1.0

1.2

1.4

1.6

Equivalence ratio Fig. 11. Comparison of predicted laminar flame speeds of n-dodecane/air at 403 K and 1 atm to the experimental data by two different methods used by Ji et al. [21] and those by Kumar and Sung [35].

fuel samples. As shown in Fig. 12, predictions are in excellent agreement with the data for these fuels. The surrogate blends for both F–T fuels give almost the same measured and predicted laminar flame speeds. This suggests that both F–T fuels should have the same flame-propagation behavior despite the differences in carbon number distributions. It also shows that laminar flamespeed measurements are not highly sensitive to differences in fuel composition when the fuels are composed mostly of n-alkanes with carbon numbers greater than 6. In addition, measured laminar flame speeds of a biomass-derived alternate jet fuel, R-8 [37], are also shown in Fig. 12. As expected, laminar flame speeds of R-8 are similar to those of F–T fuels. 4.4. Burner stabilized flames Experimental data for fuel-rich n-decane combustion in a burner stabilized flame have been obtained from Doute et al. [38]. They used gas chromatography to measure species profiles using an atmospheric pressure flame at an equivalence ratio of 1.7 (ndecane/O2/N2: 3.2/28.6/68.2 mol%) with an inlet velocity of 11.7 cm/s. The modeling of these experiments has been performed under fixed-temperature conditions, using the temperature profile reported in the paper. Doute et al. [38] estimated the uncertainty in their temperature measurements as ±5%. Figure 13 compares the calculated species profiles with the experimental data. The comparisons show good agreement for most species. For this fuel-rich condition, more CO is formed than CO2, and the model predictions

0 0

0.1

0.2

0.3

0.4

0.5

Distance (cm) Fig. 13. Comparison of predicted species profiles for a n-decane/O2/N2 burnerstabilized atmospheric pressure flame with the experimental data of Doute et al. [38]. The equivalence ratio was 1.7 with 3.2 mol% fuel with an inlet velocity of 11.7 cm/s.

follow the correct trends. The major product H2 is captured well. The peak for the major intermediate ethylene is slightly underpredicted and is shifted towards the burner surface compared to the data, but the trend is still reasonably well represented. Acetylene, which is a potential soot precursor, has a significant concentration under these fuel-rich conditions, and the model predictions of the acetylene profile agree well with the experimental data. 4.5. Extinction strain rates In two different studies Ji et al. [21,36] measured the extinction strain rates for n-dodecane/air and the two F–T fuel samples, as well as R-8 sample. These data were measured using a counterflow burner assembly (similar to that shown in Fig. 4). The experiments involve streams of fuel/air mixtures at an unburned-mixture temperature of 403 K counter-flowing against nitrogen streams of equal momentum at room temperature. We used the CHEMKIN-PRO Flame Extinction Simulator [39] for simulation of the experimental conditions. The CHEMKIN Flame Extinction Simulator performs a series of opposed-flow flame simulation with varying strain rates to find an extinction strain rate for each fuel/air mixture. Thus, extinction simulation at any one fuel/ air ratio requires hundreds of simulation of opposed-flow flame at various strain rates. This process is computationally expensive especially with the large size of a detailed reaction mechanism.

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

1200

Fuel-air nozzle at 403 K N2 nozzle at 298 K 1 atm

1000 800 600 Model, F-T surrogate (Shell GTL) Model, n-dodecane Mode, n-dodecane, initial mechanism Ji et al., n-dodecane Ji et al., Shell GTL Ji et al., S-8 Ji et al., R-8

400 200 0 0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

Fuel/Air mass ratio Fig. 14. Comparison of predicted extinction strain rates for n-dodecane/air and Shell GTL/air with the experimental data of Ji et al. [36]. The experimental data also includes S-8 and R-8 fuels [36].

Since our master kinetic mechanism is relatively large, we first generated a reduced mechanism for the extinction strain-rate calculations, making sure not to lose any accuracy in the process. The mechanism reduction was performed using methods reported previously by Naik et al. [8], in order to produce a mechanism that is accurate for flame simulation over a wide range of equivalence ratios. For these simulations, a conservatively reduced mechanism (no loss of important details and accuracy) containing 234 species and 1605 reactions was used. We also verified the accuracy by comparing to calculations using the full master mechanism for extinction strain rates at one fuel–air mixture and for laminar flame speeds calculations at several equivalence ratios. As shown in Fig. 14, lines representing the predictions do not appear particularly smooth as only a limited number of simulation points were generated due to the high computational costs of the simulations. Mixture-averaged transport properties with thermal diffusion effects were used in CHEMKIN simulations after determining that the error of this assumption was relatively small. While the use of multi-component transport properties is preferable for extinction strain-rate calculations, it is significantly more computationally expensive. However, results employing the multi-component transport formulation are expected to be within 5% of those obtained using the mixture-averaged transport formulation for the premixed fuel/air extinction cases shown here, as suggested by Ji et al. [21] for premixed flames. We note that under the more transport-dependent conditions of diffusion flames, one would expect multi-component transport to have a greater impact. Figure 14 shows that the model predictions for extinction strain rates compare well with data and differences are within the uncertainty of the experimental data, over the range of fuel/air ratios studied. Extinction strain rates for n-dodecane are similar to those of F–T fuels as well as to the bio-jet fuel R-8. Simulation results for n-dodecane using the initial source mechanism are also shown to demonstrate the improvement achieved using the current mechanism. The favorable agreement between the predictions and the experimental data, for both the extinction strain rates and the laminar flame speeds, provides confidence in the fuel-component mechanisms and in the surrogate-blending methodology used to determine the surrogate composition.

using a high-accuracy chemiluminescence analyzer, which measures NOx on a dry basis [40]. Comparisons of predicted and measured NOx for n-decane/air and n-dodecane/air flames are shown in Fig. 15. The measured NOx maximum concentration is approximately 29 ppm (on dry basis) for fuel-lean flames and 119 ppm for stoichiometric flames, with the peak location almost at the center of the burners separation distance. Both n-decane and n-dodecane flames produce very similar NOx concentrations under all conditions. The model slightly underpredicts the peak NOx concentration by 2 ppm for fuel-lean conditions and by 10 ppm for the stoichiometric conditions, which is better than 10% accuracy. The variation in measured and predicted levels of NOx is larger near burners for stoichiometric conditions, for which we have no explanation as of yet. The NOx data were obtained for two F–T fuel samples as well as for a biomass-derived alternate jet fuel, R-8 [37]. Since detailed fuel analysis was not available for R-8 fuel, it was not possible to generate an optimized surrogate blend. Figure 16 shows the comparison of the measured NOx concentrations in the fuel/air flames (fuel to air mass ratio of 0.0532 at 403 K under a strain rate of approximately 166 s 1), to those predicted by the model using appropriate surrogates for two F–T fuel samples (Table 3). As seen in the figure, measured NOx emissions from S-8 and Shell GTL fuels are very similar, and this trend is also accurately captured by the model. NOx emission from R-8 fuel is slightly less than that from

140 120

Data, n-decane

Fuel/air at 403 K, 1 atm Strain rate 166 1/s

Data, n-dodecane Model, n-decane

100

NOx (ppm)

Extinction Strain Rate (1/s)

1400

Data, n-dodecane Model, n-decane Model, n-dodecane

60 φ =0.8

20 0 0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

1.10

1.20

Distance from the bottom burner (cm) Fig. 15. Comparison of predicted and measured (this work) NOx levels for counterflow n-decane/air and n-dodecane/air vs. N2 flames. Fuel/air is injected at 403 K and 1 atm with mass equivalence ratios 0.8 and 1 at global strain rate of 166 1/s.

35 Fuel/air 403 K, 1 atm, Strain rate 166 1/s Fuel/air mass ratio 5.32%

30 25 20 15 10

Model, S-8 Model, Shell GTL Data, S-8 Data, Shell GTL Data, R-8

0 0.3

NOx emissions were measured in this work using a counterflow burner assembly with premixed fuel/air mixtures from the heated bottom burner counterflow against nitrogen at room temperature from the top burner (Fig. 4). NOx concentrations were measured

Data, n-decane

80

5

4.6. NOx emissions

Model, n-dodecane

φ =1

40

NOx (ppm)

442

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Distance from the bottom burner (cm) Fig. 16. Comparison of predicted and measured (this work) NOx from premixed F–T and bio-jet fuels vs. N2 in counterflow flames. Fuel/air is injected at 403 K and 1 atm with mass ratio of 5.32% at global strain rate is 166 1/s.

443

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

the F–T fuels, but still within the uncertainty range of the measurement. Under all conditions, the majority of NOx is in the form of nitric oxide (NO), as predicted by the model.

5. Discussion To gain some insight into the detailed mechanism for the fuels surrogates, reaction path and sensitivity analysis has been performed for several fundamental experimental configurations using the actual alternative jet fuels. Reaction path analyses revealed the usual paths expected for hydrocarbon combustion at high-temperatures, as shown in Fig. 17. Fuel consumption under all experimental conditions including autoignition and flames were dominated by H atom abstraction by either OH or H radicals. Further reaction involves abstraction followed by b-scission steps that break down larger molecules into smaller hydrocarbons including olefins and radicals. Sensitivity analyses have been performed using CHEMKIN-PRO software [39] to understand the relative importance of various pathways. Sensitivity coefficients presented in this work are the first-order normalized sensitivity coefficients in the form of logarithmic derivatives as calculated by CHEMKIN-PRO software [39]. Analyses under three different conditions are discussed: (1) temperature sensitivity analysis during autoignition of n-dodecane at 1100 K which is the lower end of the temperature range for the jet engine applications, for conditions corresponding to those in Fig. 7, (2) sensitivity analysis for laminar flame speeds of the S-8 surrogate blend for conditions corresponding to those used in Fig. 12, and (3) sensitivity analysis for NO from the S-8 flames under the conditions corresponding to those used for Fig. 16. RH

+H/OH -H2 /H2O R•

+O2 -HO2 Olefin + R’•

Abstraction and β scission Smaller olefin + R’’•

CH3•,C2H5•,etc Fig. 17. Overall reaction scheme for high-temperature oxidation of F–T fuel surrogates.

Temperature sensitivity coefficients for autoignition of n-dodecane with air at 20 atm and O2/Ar at 16 atm under shock-tube conditions of Fig. 7, at an equivalence ratio of 0.5, 1100 K, and 75% fuel conversion, are shown in Fig. 18. We can see the similarities and differences in sensitivity under these two conditions. For fuel/O2/Ar mixtures, results are most sensitive to chain branching reaction Rxn 1, whereas for the fuel/air mixture results are most sensitive to the pressure-dependent dissociation reaction Rxn 2. In general, reactions involving HO2 radicals, including their self-reaction and abstraction reactions, are more important for fuel/air oxidation than for fuel/O2/Ar under comparable conditions. Reactions of smaller hydrocarbons such as Rxn 3 and Rxn 4 are important under both conditions. Rate coefficients of these reactions are listed in Table 4. Rxn Rxn Rxn Rxn Rxn Rxn Rxn

1: 2: 3: 4: 5: 6: 7:

H + O2 = O + OH H2O2 (+M) = 2 OH (+M) CH3 + HO2 = CH3O + OH CH3 + HO2 = CH4 + O2 CO + OH = CO2 + H HCO + M = CO + H + M N + NO = N2 + O

As shown in Fig. 19, the reactions of hydrogen and the smaller hydrocarbons C0–C3 are most important for laminar flame speeds for the S-8 surrogate blend. The highest sensitivity coefficients are for Rxn 1, Rxn 5, and Rxn 6. As listed in Table 4, the rate constants for Rxn 6 have been adjusted to improve predictions over a broad range of validation conditions. The variation is a factor of two higher than that recommended by Friedrichs et al. [41], but still within the uncertainty and scatter among various experimental data. For this reaction, experimentally measured rate constants vary by a factor of three to eight at different temperatures [23]. No reactions involving heavier hydrocarbons appear to cause sensitivity for the laminar flame conditions. Similar conclusions were also reached by Ji et al. [21] for several heavier hydrocarbon flames, where they concluded that the chemistry of smaller hydrocarbons plays an important role even for larger hydrocarbon and surrogate flames. Consistent with this, the kinetics of the core mechanism improvements in this work actually resulted in the improved results for the higher hydrocarbons. Under the S-8/air vs. N2 counterflow flame configuration used for the NOx measurements, NO sensitivity analysis presented in Fig. 20 shows that the most important reactions for NOx formation were also found to be important for the laminar flame speeds. Only one additional reaction Rxn 7 was found to be the most sensitive for NO formation. This reaction is part of the thermal-NOx formation mechanism, which is important for NO formation under the flame conditions under discussion. Although thermal-NOx pathway dominates under conditions in flames, other pathways included in

2ch3(+M)<=>c2h6(+M) h2o2(+M)<=>2oh(+M) ch3+ho2<=>ch3o+oh

h+o2<=>o+oh

h+o2<=>o+oh

co+oh<=>co2+h

nc12h26+ho2<=>c12h25-3+h2o2

hco+M<=>co+h+M

nc12h26+ho2<=>c12h25-5+h2o2

c2h3+o2<=>ch2cho+o

c2h3+o2<=>ch2cho+o

ch2(s)+o2=>h+oh+co

c2h4+oh<=>c2h3+h2o c2h3+o2<=>ch2o+hco ch3+ho2<=>ch4+o2

h+o2(+M)<=>ho2(+M) c3h5-a+h(+M)<=>c3h6(+M)

n-dodecane/air, 20 atm n-dodecane/O2/Ar, 16 atm

2ho2<=>h2o2+o2

-0.02

ch3+h(+M)<=>ch4(+M)

Autoignition of n-dodecane at phi 0.5, 75% conversion

-0.01

0

0.01

0.02

0.03

hco+h<=>co+h2

0.04

Temperature sensitivity coefficient

S-8, phi1

h+oh+M<=>h2o+M -0.1

0

0.1

0.2

0.3

0.4

Laminar flame speed sensitivity coefficient Fig. 18. Temperature sensitivity analysis for autoignition of n-dodecane with air and O2/Ar under shock-tube conditions used for simulation results shown in Fig. 7 at equivalence ratio of 0.5, 1100 K, and 75% fuel conversion.

Fig. 19. Sensitivity analysis for the laminar flame speeds of the S-8 surrogate blend for 403 K, 1 atm, and an equivalence ratio of 1.

444

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445

6. Summary

n+no<=>n2+o co+oh<=>co2+h h+o2<=>o+oh ch+n2<=>hcn+n hco+M<=>co+h+M ch+o2<=>hco+o h+oh+M<=>h2o+M h+o2(+M)<=>ho2(+M) -0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

NO sensitivity coefficient Fig. 20. Sensitivity analysis for nitric oxide in the S-8/air vs. N2 counterflow flames under conditions used for simulation results shown in Fig. 16.

the comprehensive NOx mechanism also play an important role under conditions tested for smaller hydrocarbon validation results reported in Supplementary material. In addition, we used CHEMKINPRO [39] Uncertainty Analysis feature to quantify the effect of uncertainty in the measurement of nozzle velocities and deviations in nozzle temperatures. Variation in nozzle velocities by 5% and deviations in nozzle temperature by 10 K can have sufficient effect on NOx values to explain the difference between predictions and measured values away from the peak, as shown in Fig. 21. Based on this uncertainty analysis, current predictions are considered to be in reasonably good agreement with the measured values. Successful validation of the model performed not only under flame conditions, but also under extended temperature, pressure, and equivalence ratio conditions, supports that the proposed three-component surrogates and its detailed reaction mechanism can capture the combustion and emissions properties of alternate jet fuels. The self-consistent and rate-rules based approach of the mechanism also increase the likelihood of accurate predictions even under conditions where the mechanism has not been explicitly validated; in other words, the mechanism can be used as a predictive tool. The main limitation of the mechanism is that it is applicable to a high-temperature regime above 1100 K, since low-temperature reaction pathways were omitted to keep the size of the detailed mechanism smaller for faster computational time and to facilitate use in CFD applications after mechanism reduction. As demonstrated by Naik et al. [8] the mechanism can be reduced by more than 90% by using automated mechanism reduction methods. While the PAH part of the detailed mechanism has not been validated here, in the future we plan to validate and improve the PAH mechanism assembled in this work.

40 Model, n-decane

Fuel/air at 403 K, 1 atm Strain rate 166 1/s f = 0.8

Model, n-dodecane

NOx (ppm)

30

Data, n-decane Data, n-dodecane Model, 10 K Std. dev. in nozzle temperature Model, 5% Std. dev. in nozzle velocities

20

10

0 0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

Distance from the bottom burner (cm) Fig. 21. Predicted effect of uncertainties in nozzle velocities (5% standard deviation) and nozzle temperatures (10 K standard deviation) on NOx levels for n-decane/air and n-dodecane/air flame at equivalence ratio of 0.8 under conditions used for Fig. 15. Open symbols represent predicted uncertainties.

Three-component F–T fuel surrogate blends containing n-dodecane, n-decane, and iso-octane have been proposed and their compositions have been optimized to represent various properties of fuels from two different sources. The same surrogates for the F–T fuel samples were also used for comparing the predictions to the data for a bio-derived jet-fuel sample, R-8. A self-consistent detailed chemical kinetic mechanism for modeling the surrogate components’ combustion and emissions behavior has been assembled from several component mechanisms built on rate-rules, and has been improved to accurately predict fundamental properties, including laminar flame speeds, and extinction strain rates. The mechanism has been validated against a variety of fundamental experimental data for individual components over a broad range of conditions including autoignition time, autoignition temperature, laminar flame speeds, extinction strain rates, and NOx emissions. The main deficiency of the current validation is the lack of validation for sooting or near-sooting conditions. Model predictions were compared to experimental data for the actual fuels, showing the effectiveness of the surrogate-blending approach, as well as the accuracy of the detailed kinetic mechanisms. In addition to using fundamental experimental data from the literature used for validation in this work, NOx measurements in stretched premixed laminar flames for a few fuels, including the F–T and bio-jet fuels were performed as part of this study. The combination of targeted experiments and the literature data provide good coverage of the mechanism capabilities for the high-temperature conditions of jet-engine combustion. Acknowledgments The authors acknowledge the financial support of NASA, under contract number NNC07CB45C, for this work. The authors thank Prof. Fokion N. Egolfopoulos for his guidance and technical discussions. The authors also thank Dr. Charles K. Westbrook for helpful discussions in identifying the initial surrogate components.

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2010.09.016. References [1] T. Kalnes, T. Marker, D.R. Shonnard, Int. J. Chem. React. Eng. 5 (2007) A48. [2] G.W. Huber, Report on Breaking the Chemical and Engineering Barriers to Lignocellulosic Biofuels: Next Generation Hydrocarbon Biorefineries; University of Massachusetts, Amherst. , 2007. [3] C.A. Moses, Report on Comparative Evaluation of Semi-Synthetic Jet Fuels; Coordinating Research Council, Project No. AV-2-04a, September, 2008. [4] T. Edwards, D. Minus, W. Harrison, E. Corporan, AIAA-2004-3885, 2004. [5] C.V. Naik, K.V. Puduppakkam, E. Meeks, SAE Technical Papers 2010-01-0541, 2010. [6] R.H. Perry, D.W. Green, J.O. Maloney, in: Perry’s Chemical Engineers’ Handbook, seventh ed., McGraw-Hill, 1997. [7] N.A. Slavinskaya, A. Zizin, M. Aigner, in: Proc. ASME Turbo Expo GT200960012, 2009. [8] C.V. Naik, K.V. Puduppakkam, A. Modak, C. Wang, E. Meeks, in: Proc. ASME Turbo Expo GT2010-23709, 2010. [9] C.K. Westbrook, W. Pitz, O. Herbinet, H.J. Curran, E.J. Silke, Combust. Flame 156 (2009) 181–191. [10] C.V. Naik, A.M. Dean, Proc. Combust. Inst. 32 (1) (2009) 437–443. [11] H.J. Curran, P. Gaffuri, W.J. Pitz, C.K. Westbrook, Combust. Flame 129 (2002) 253–280. [12] Reaction Design CHEMKIN-PRO, 13082, Reaction Design, 2008. [13] H.J. Curran, P. Gaffuri, W.J. Pitz, C.K. Westbrook, Combust. Flame 114 (1998) 149–177. [14] K.G. Joback, R.C. Reid, Chem. Eng. Commun. 57 (1987) 233–243.

C.V. Naik et al. / Combustion and Flame 158 (2011) 434–445 [15] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M. Goldenberg, C.T. Bowman, R.K. Hanson, S. Song, J. William, C. Gardiner, V. V. Lissianski, Z. Qin GRI-Mech 3.0. . [16] P. Dagaut, P. Glarborg, M.U. Alzueta, Prog. Energy Combust. Sci. 34 (2008) 1– 46. [17] C.L. Rasmussen, A.E. Rasmussen, P. Glarborg, Combust. Flame 154 (3) (2008) 529–545. [18] J. Appel, H. Bockhorn, M. Frenklach, Combust. Flame 121 (2000) 122–136. [19] M.S. Skjoth-Rasmussen, P. Glarborg, M. Ostberg, J.T. Johannessen, H. Livbjerg, A.D. Jensen, T.S. Christensen, Combust. Flame 136 (2004) 91–128. [20] H.R. Zhang, E.G. Eddings, A.F. Sarofim, C.K. Westbrook, Proc. Combust. Inst. 32 (2009) 377–385. [21] C. Ji, E. Dames, Y.L. Wang, H. Wang, F.N. Egolfopoulos, Combust. Flame 157 (2009) 277–287. [22] C.V. Naik, E. Meeks, S. Drennan, Detailed Chemical Kinetic Mechanism for Methane and Ethane Combustion Including NOx Emissions for Gas Turbines Applications, Fall Technical Meeting of the Western States Section of the Combustion Institute Irvine, CA, 2009. [23] NIST Chemical Kinetics Database, Standard Reference Database 17, 7.0 (Web Version) , 2010. [24] H. Wang, X. You, A.V. Joshi, S.G. Davis, A. Laskin, F. Egolfopoulos, C.K. Law, USC Mech Version II. High-Temperature Combustion Reaction Model of H2/CO/C1– C4 Compounds. (May 2007). [25] A.T. Holley, Y. Dong, M.G. Andac, F.N. Egolfopoulos, T. Edwards, Proc. Combust. Inst. 31 (2006) 1205–1213. [26] D.F. Davidson, M.A. Oehlschlaeger, J.T. Herbon, R.K. Hanson, Proc. Combust. Inst. 29 (2002) 1295–1301. [27] D.J. Vermeer, J.W. Meyer, A.K. Oppenheim, Combust. Flame 18 (1972) 327– 336. [28] U. Pfahl, K. Fieweger, G. Adomeit, Proc. Combust. Inst. 26 (1996) 781–789. [29] V.P. Zhukov, V.A. Sechenov, A.Y. Starikovskii, Combust. Flame 153 (2008) 130– 136.

445

[30] S.S. Vasu, D.F. Davidson, Z. Hong, V. Vasudevan, R.K. Hanson, Proc. Combust. Inst. 32 (2009) 173–180. [31] T. Bieleveld, A. Frassoldati, A. Cuoci, T. Faravelli, E. Ranzi, U. Niemann, K. Seshadri, Proc. Combust. Inst. 32 (2009) 493–500. [32] S.G. Davis, C.K. Law, Proc. Combust. Inst. 27 (1998) 521–527. [33] Y. Huang, C.J. Sung, J.A. Eng, Combust. Flame 139 (2004) 239–251. [34] K. Kumar, J.E. Freeh, C.J. Sung, Y. Huang, J. Propul. Power 23 (2) (2007) 430. [35] K. Kumar, C.J. Sung, Combust. Flame 151 (1–2) (2007) 209–224. [36] C. Ji, Q. Feng, T.T. Tsotsis, F.N. Egolfopoulos, J. Propul. Power, 2010, submitted for publication. [37] T. Edwards, US Air Force Alternative Fuel Efforts: Fischer–Tropsch and Beyond, Fuels Research Review Multi Agency Coordination Committee for Combustion Research, Los Angeles, 2009. [38] C. Doute, J.-L. Delfau, C. Vovelle, Combust. Sci. Technol. 130 (1997) 269–313. [39] Reaction Design CHEMKIN-PRO, Pre-release; Reaction Design, 2010. [40] Q. Feng, Y.L. Wang, F.N. Egolfopoulos, T.T. Tsotsis, Ind. Eng. Chem. Res. (2010), doi:10.1021/ie100481q. [41] G. Friedrichs, J.T. Herbon, D.F. Davidson, R.K. Hanson, Phys. Chem. Chem. Phys. 4 (2002) 5778–5788. [42] D.L. Baulch, C.J. Cobos, R.A. Cox, C. Esser, P. Frank, T. Just, J.A. Kerr, M.J. Pilling, J. Troe, R.W. Walker, J. Warnatz, J. Phys. Chem. Ref. Data 21 (1992) 411–429. [43] A.V. Joshi, H. Wang, Int. J. Chem. Kinet. 38 (2006) 57–73. [44] J. Troe, Proc. Combust. Inst. 28 (2000) 1463–1469. [45] X. You, H. Wang, E. Goos, C.J. Sung, S.J. Klippenstein, J. Phys. Chem. A 111 (19) (2007) 4031–4042. [46] J.A. Miller, S.J. Klippenstein, Phys. Chem. Chem. Phys. 6 (2004) 1192–1202. [47] W. Tsang, J.T. Herron, J. Phys. Chem. Ref. Data 20 (1991) 609–663. [48] W. Tsang, R.F. Hampson, J. Phys. Chem. Ref. Data 15 (1986) 1087–1279. [49] D.F. Davidson, B.M. Gauthier, R.K. Hanson, Proc. Combust. Inst. 30 (2005) 1175–1182. [50] C. Doute, J.-L. Delfau, R. Akrich, C. Vovelle, Combust. Sci Technol. 106 (1995) 327–344.