MEASUREMENTS OF IGNITION TIMES, OH TIME-HISTORIES, AND REACTION RATES IN JET FUEL AND SURROGATE OXIDATION SYSTEMS

Transcription

1 MEASUREMENTS OF IGNITION TIMES, OH TIME-HISTORIES, AND REACTION RATES IN JET FUEL AND SURROGATE OXIDATION SYSTEMS A DISSERTATION SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Subith S. Vasu August 2010

2 2010 by Subith Vasu Sumathi. All Rights Reserved. Re-distributed by Stanford University under license with the author. This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. This dissertation is online at: ii

3 I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Ronald Hanson, Primary Adviser I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Craig Bowman I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. David Golden Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives. iii

4 Abstract Fossil-based hydrocarbon fuels account for over 80% of the primary energy consumed in the world - it is still expected to be about 70% in year and nearly 60% of that amount is used in the transport sector. The basis for globalization is transportation and a driving force has been the growth in global air traffic. The current climate crisis magnifies the need for improving the performance of jet engines by introducing scientific designs in which the use of chemical kinetics will be essential and critical for better performance and reducing pollutant emissions. Most aviation fuels are jet fuels originating from crude oil and there are major gaps in our knowledge of the hightemperature chemistry of real liquid carbon-based fuels. There is a critical need for experimental kinetic databases that can be used for the validation and refinement of jet fuel surrogate mechanisms. To fill this need, experiments were performed using shock tube and laser absorption methods to investigate jet fuel and surrogate oxidation systems under engine-relevant conditions. Ignition times and OH species time-histories were measured and low-uncertainty measurements of the reactions of OH with several stable intermediates were carried out. The work presented in this study can be broken into three categories: 1) jet fuel oxidation, 2) surrogate oxidation, and 3) OH radical reactions with several stable combustion intermediates. Ignition delay times were measured for gas-phase jet fuel oxidation (Jet-A and JP-8) in air behind reflected shock waves in a heated high-pressure shock tube. Initial reflected shock conditions were as follows: temperatures of K, pressures of atm, equivalence ratios ( ) of 0.5 and 1, and oxygen concentrations of 10 and 21 % in synthetic air. Ignition delay times were measured using sidewall pressure and OH* emission at 306 nm. The new experimental results were modeled using several kinetic mechanisms using various jet fuel surrogate mixtures. Normal and cyclo alkanes are the two most important chemical classes found in jet fuels. Ignition delay time experiments were conducted during high-pressure oxidation of two commonly used representative components for normal and cyclo alkanes in jet fuel surrogates, i.e., n-dodecane and methylcyclohexane (MCH), respectively. Fuel/air ignition was studied for the following shock conditions: temperatures of K, pressures of 17- iv

5 50 atm, s of 0.5 and 1. OH concentration time-histories during high-pressure n-dodecane, n-heptane and MCH oxidation were measured behind reflected shock waves in a heated, high-pressure shock tube. Experimental conditions covered temperatures of 1121 to 1422 K, pressures of atm, and initial fuel concentrations of 500 to 1000 ppm (by volume), and an equivalence ratio of 0.5 with O 2 as the oxidizer in argon as the bath gas. OH concentrations were measured using narrow-linewidth ring-dye laser absorption near the R- branchhead of the OH A-X (0,0) system at nm. Detailed comparisons of these data with the predictions of various kinetic mechanisms were made. Sensitivity and pathway analyses for these reference fuel components were performed, leading to reaction rate recommendations with improved model performance. Reactions of OH radical with two alkenes (ethylene and propene) and a diene (1,3- butadiene) were studied behind reflected shock waves. Measurements were conducted in the range of temperatures from K and pressures from atm for three initial concentrations of fuels (500ppm, 751.1ppm and 1000ppm). OH radicals were produced by shock-heating tert-butyl hydroperoxide, (CH 3 ) 3 - CO-OH, and monitored by narrow-line width ring dye laser absorption of the well characterized R 1 (5) line of the OH A-X (0, 0) band near nm. OH time-histories were modeled by using a modified oxidation mechanism and rate constants for the reactions of OH with ethylene, propene, and 1,3- butadiene were extracted by matching modeled and measured OH concentration time histories in the reflected shock region. Detailed error analyses yielded an uncertainty estimate of ± 22.8% (OH+ethylene at 1201 K), ±16.5% (OH+propene at 1136 K), and ± 13% (OH+1,3-butadiene at 1200K). Canonical and variational transition state theory calculations using recent ab initio results gave excellent agreement with our experimental measurements and data outside our range and hence the resulting expressions can be used directly in combustion models. In the current studies, a rate measurement for the decomposition of TBHP has been obtained in the range K using both incident and reflected OH data. v

6 Acknowledgements I owe a great deal of gratitude to my advisor, Professor Ronald Hanson, for the opportunities, support, encouragement and guidance he has provided me during my time at Stanford. His work ethic, creative ideas, and critical thinking are unmatchable in this world and I feel like I still have a lot to learn from him. I would like to thank my reading committee, Professors David Golden and Craig Bowman for their valuable suggestions regarding this thesis. I have been fortunate enough to work with Dr. David Davidson who is not only an outstanding scientific researcher, but also an incredible advisor regarding life. I am particularly grateful to Matt Oehlschlaeger, Venkatesh Vasudevan and Rob Cook for teaching me early in my research work. I would like to thank Prof. Heinz Pitsch and Dr. Jay Jeffries for their support and help. I would like to give special thanks to all my friends and colleagues in the Hanson research group (current and former). I am also grateful to everyone and my friends who have encouraged and helped me develop as a researcher. Finally, I would like to acknowledge the constant support my family has been providing me. This research was sponsored by the U.S. Army Research Office (ARO) and the U.S. Department of Energy. vi

7 Contents Abstract... iv Acknowledgements... vi Contents... vii List of Tables... x List of Figures... xii Chapter 1 : Introduction Motivation Background Jet Fuel Ignition Times Surrogate Ignition Times High-pressure OH Time-histories During Surrogate Oxidation Reaction of OH with Alkenes OH+1,3-Butadiene Scope and Organization of Thesis Chapter 2 : Method Shock Tube Facilities High-pressure Shock Tube Low-pressure Shock Tube OH Laser Absorption Diagnostics Modeling Shock Tube Data Chapter 3 : High-pressure Jet Fuel Ignition Times Jet Fuel/Air Mixture Preparation Measuring Ignition Delay Times Jet Fuel Ignition: Results and Discussion Comparison with Previous Measurements Comparison of Different Fuel Types Pressure Scaling Jet Fuel Surrogate Modeling Prediction vii

8 3.3.5 Effect of Variation of Equivalence Ratio and Oxygen Concentration Low-temperature (NTC Region) Ignition Delay Time Data Chapter 4 : High-pressure Single-Component Surrogate Ignition Times Experimental Method n-dodecane/air Ignition: Results and Discussion High-pressure n-dodecane Ignition Data Kinetic Modeling Prediction Effect of Pressure-time Histories Comparison of n-dodecane Ignition Times with Jet-A and Other n- Alkanes MCH/Air Ignition: Results and Discussion High-pressure MCH Ignition Data Comparison of Various Cyclo-alkanes Kinetic Modeling Effect of Pressure-time Histories Comparison of Jet Fuel and Surrogate Component Ignition Chapter 5 : OH Time-histories During Surrogate Oxidation at High-Pressure Experimental Method Kinetic Modeling Details OH Time-Histories Results n-heptane Results MCH Results n-dodecane Results Discussion n-heptane Kinetic Analysis MCH Kinetic Analysis n-dodecane Kinetic Analysis Chapter 6 : Alkenes Reaction With OH: OH+C 2 H 4 and OH+C 3 H Method Kinetic Measurements OH+C 2 H 4, k viii

9 6.2.2 OH+C 3 H 6, k TBHP Decomposition, k Comparison with Previous Data Theoretical Calculations of OH+C 2 H 4 and OH+C 3 H Chapter 7 : 1,3-Butadiene+OH Products Method Kinetic Measurements TST Calculations Chapter 8 : Conclusions Summary of Results Jet Fuel Oxidation Surrogate Fuels Oxidation Reactions of OH with Alkenes OH+1,3-Butadiene Archival Publications Recommendations for Future Work Multi-species Measurements in Multi-component Mixtures Measurements of Reactions in the NTC Regime Using OH Lasers Kinetics of Biobutanol Appendix A: Modes of Ignition in High-Pressure Fuel/Air Mixture Ignition Appendix B: Influence of Impurities, Particles, and Wall Effects on Ignition Appendix C: Shock Tube Boundary Layer and Facility Effects on Ignition Appendix D: TST Theory and Calculations Input Files to Multiwell Code References ix

10 List of Tables Table 1.1 High-temperature OH+ethylene and OH+propene studies in literature Table 3.1 Summary of Jet-A/air ignition data at phi=1.0. Jet-A Composite Blend #04POSF4658. (Dr. = driver gas mixture; V shock = incident shock velocity at the endwall) Table 3.2 Summary of jet fuel ignition data indicating variation with fuel type (jet-a and JP-8), equivalence ratio, and fuel sample. (Dr. denotes driver gas mixture; V shock is the incident shock velocity at the endwall) Table 3.3 Jet fuel surrogate mechanisms and their associated surrogate mixtures Table 3.4 Jet fuel surrogate mixtures used with the Ranzi [48] and the Zhang et al. [19] mechanisms Table 4.1 Summary of current shock tube experimental results in n-dodecane/air, phi= Table 4.2 Summary of current shock tube experimental results in n-dodecane/air, phi= Table 4.3 Summary of current high-pressure ignition time results in MCH/air ( =1.0). Mixture: MCH=1.96%, O 2 =20.60%, N 2 =77.44%. V shock is the incident shock velocity at the endwall Table 5.1 Mechanisms used in this study of surrogate components Table 5.2 Summary of current high-pressure OH absorption experiments in MCH and n- heptane. a Sat.= saturated signal (transmission~0) and in this case, ign was obtained from raw absorption signals instead of X OH profiles. *Off=Offline absorption measurement. b NL=No laser was used. ign was obtained from pressure traces as the midpoint of pressure jump during ignition for both Off and NL cases Table 5.3 Summary of OH absorption data (fuel=n-dodecane). a Sat., saturated signal;transmission~ Table 6.1 OH+ethylene rate coefficient data Table 6.2 Uncertainty analysis for OH+alkenes reactions. OH+ethylene (k 1 ): initial incident shock conditions: 1201 K, 1.99 atm, and 500 ppm ethylene/ar. x

11 OH+propene (k 2 ): initial incident shock conditions: 1136 K, 2.02 atm, and 299 ppm ethylene/ar Table 6.3 OH+propene rate coefficient data Table 6.4 TBHP decomposition rate data. Superscripts denotes mixtures: a 500ppm ethylene; b 1000ppm ethylene; c 300 ppm propene Table 7.1 Overall OH+1,3-butadiene rate coefficient, k 1, data Table 7.2 Uncertainty analysis for overall OH+1,3-butadiene rate coefficient, k Table 7.3 Energies of the products and the transition states in kcal mol -1 units relative to the reactants trans-1,3-butadiene + OH, without and with zero-point energy. Q1 represents the Q1 diagnostic of the QCISD(T)/cc-pVQZ calculations at the B3LYP/ G(d,p) geometry. The values are very close to the ones obtained at the MP2 geometry (not shown) Table 7.4 Energies The equilibrium constant and K eff of the 1,3-butadiene + OH CH 2 CHCHCH 2 OH reaction. For details, see text Table 8.1 Laser-based quantitative absorption diagnostics developed in our lab [194]. 131 xi

12 List of Figures Figure 1.1 Comparison of rate constants for C 2 H 4 +OH used in several current mechanisms. Variations of a factor of ~5.5 are evident at high temperatures near 1000 K. Shown are rates used in JetSurF 1.0 [46], Westbrook et al. [47], Ranzi [48], Galway natural gas mechanism III [51], and Chaos et al. [52] Figure 2.1 Experimental set-up for HPST experiments Figure 3.1 Example JP-8/air ignition delay time data (driver gas: He). Initial reflected shock conditions, 1019 K, 22.2 atm, = Figure 3.2 Ignition delay times including previous data for Jet-A/air for phi=1. Solid circles: current work for Jet-A POSF4658; open circles: Dean et al. [9]; open squares: current work for Dean et al. [9] Jet-A fuel Figure 3.3 Ignition delay times for Jet-A/air and JP-8/air, phi= Figure 3.4 Pressure scaling of (phi=1) τ ign : left- Jet-A/air; right- Jet-A/air and JP-8/air. 26 Figure 3.5 Jet fuel ignition delay times for phi=1 (data from Figure 3.3), and mechanisms prediction. The Violi #3 surrogate was used with the Ranzi and the Zhang et al. mechanisms. See text for details Figure 3.6 Jet fuel ignition delay times for phi=1, and the Ranzi mechanism predictions for different surrogate fuel mixtures listed in Table Figure 3.7 Jet fuel ignition fuel ignition delay times: the effect of phi and X O Figure 3.8 Left: low-temperature pressure and emission traces (Driver gas: He 70%, N 2 30 %). Right: ignition delay times including NTC region data, (Jet-A/air, phi=1).. 33 Figure 4.1 Example ignition data from HPST. Left: n-dodecane/air. Right: MCH/air. 40 Figure 4.2 n-dodecane/air τ ign data (20atm) and predictions for =0.5 (left) and1.0 (right) Figure 4.3 Left: n-dodecane /air τ ign results for phi=1.0. CHEMSHOCK modeling was conducted using the You et al. mechanism with an extreme-case linear pressure rise (dp 5 /dt) of 10% per milli-second. Right: Measured n-dodecane/air pressure-time histories near 20 atm, phi= xii

13 Figure 4.4 Comparison of high-pressure n-dodecane ignition delay times at 20 atm. Left: with Jet-A/air at two equivalence ratios. Right: with n-heptane/air (current work) at =1.0. Dashed lines are fit through data. Solid lines are Westbrook et al. [47] (LLNL) modeling prediction Figure 4.5 Left: High-pressure MCH/air ignition delay time results ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%). High-pressure data scaled to 20, or 45 atm using τ ign ~ P Low-pressure data scaled to 1.5atm using Vasu et al. [121] correlation ( =1.0, MCH=1.962 %). Modeling shows τ ign predictions at 20 atm. Right: High-pressure MCH/air ignition delay time results near 45 atm ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%) and pressure scaling. Data (solid symbols) scaled to 45 atm using τ ign ~ P Grey solid line is a fit through data. Constant U,V modeling results at 20 and 45 atm are shown using Orme et al., Ranzi et al., and Pitz et al. mechanisms Figure 4.6 Current high-pressure MCH/air ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%) τ ign results. Left: Current data scaled to 20, or 45 atm using τ ign ~ P (- 0.87). RCM data (unheated) from Pitz et al. [22]. Lines are constant U,V modeling predictions using the Pitz et al. [22] mechanism: 1) 10 atm; 2) 20 atm; and 3) 45 atm. Right: HPST data scaled to 20, or 45 atm using τ ign ~ P (-0.87) ; RCM data (unheated) from Pitz et al. [22]; Solid lines are fit through data at respective pressures. Lowpressure data scaled to 1.5atm using Vasu et al. [121] correlation ( =1.0, MCH=1.962 %) Figure 4.7 High-pressure cyclo-alkanes/air ( =1.0) τ ign results. Mixtures: MCH=1.96%, O 2 =20.60%, N 2 =77.44%; cyclo-hexane=2.28%, O 2 =20.53%, N 2 =77.19%; cyclopentane=2.72%, O 2 =20.44%, N 2 =76.84%. Current MCH data scaled to 45 atm using τ ign ~ P (-0.87). Cyclo-hexane (scaled using τ ign ~ P -1.1 and cyclo-pentane (scaled using τ ign ~ P -0.9 ) data from Daley et al. [124]. Solid lines are fit through data Figure 4.8 Left: Measured and predicted (Ranzi et al. [48], Pitz et al. [22]) P(t) histories. Right: Computed pressure-time histories for MCH/air ignition at 20atm, =1.0, Pitz et al. [22] mechanism Figure 4.9 HPST pressure-time histories for MCH/air ( =1.0) ignition. Left: near 45atm. Right: near 20atm xiii

14 Figure 4.10 High-pressure ignition times data from HPST in Jet-A and major singlecomponent surrogate fuels at similar conditions ( =1.0). HPST data scaled to 20 atm using respective pressure scaling for individual fuels (see text). Solid lines are fit through data at 20 atm: iso-octane and toluene data from Davidson et al. [115] and Vasu et al. [143], respectively; n-dodecane, Jet-A, MCH, and n-heptane are from current work Figure 5.1 High-pressure OH absorption data for n-heptane; initial X fuel =1000ppm, X O2 =0.022, X Ar =0.977, =0.5. Data 1: 1271K, 15atm; data 2: 1236K, 15.28atm; data 3: 1230K, 15.81atm; data 4: 1121K, 14.1atm Figure 5.2 High-pressure OH absorption profile data and modeling predictions for n- heptane; initial X fuel =1000ppm, X O2 =0.022, X Ar =0.977, =0.5, 1271K, 15atm. 1: Current experiment, 2: Curran et al. [149], 3: Seiser et al. [150], 4: Tsang [151], 5: Patel et al. [153], 6: Ranzi et al. [48], 7: SanDiego [154], 8: Golovichev [152], 9: Gokulakrishnan et al. [157], 10: Chaos et al. [52], 11: Biet et al. [156], 12: You et al. [118] Figure 5.3 (A): OH absorption data for MCH; initial X MCH =1000 ppm, X O2 =0.021, X Ar =0.978, =0.5. Data 1: 1285 K, atm; data 2: 1269 K, 15.8 atm; data 3: 1213 K, atm; data 4: 1205 K, atm. (B): OH absorption data for MCH; initial X MCH =750 ppm, X O2 = , X Ar =0.9835, =0.5. Data 1: 1304 K, atm; data 2: 1303K, atm; data 3: 1266 K, atm Figure 5.4 High-pressure OH absorption data and modeling predictions (using Ranzi [48], Orme et al. [36], Pitz et al. [22] mechanisms) for MCH/O 2 /Ar. Initial X MCH =1000 ppm, X O2 =0.021, X Ar =0.978, 1262K, atm, = Figure 5.5 High-pressure OH absorption data: n-dodecane; initial X dodecane =1000ppm, O 2, Ar; =0.5. Left: Data 1: 1422K, 15.5atm; data 2: 1230K, 16.73atm; data 3: 1217K, 16.07atm; data 4: 1196K, 15.77atm; data 5: 1158K, 15.19atm. Right: Comparison of measured and modeled OH time-histories at 1217 K, 16.1 atm Figure 5.6 High-pressure OH data variation with temperature. Initial reflected shock conditions: 16 atm, 1000 ppm n-dodecane/o 2 /Ar, = 0.5. Left: peak X OH vs T 5. Solid line is a linear fit through data. Right: High-temperature, low-concentration xiv

15 ignition delay times in n-dodecane. Solid black line is least-squares fit through data Figure 5.7 Major oxidation pathways prediction using the Chaos et al. [52] mechanism integrated ROP approach. X heptane =1000ppm, X O2 =0.022, balance=ar. 1275K, 16atm, =0.5. Details of molecular structures and pathways can be found in [52] Figure 5.8 OH sensitivity for n-heptane oxidation using the Chaos et al. [52] mechanism. X heptane =1000ppm, X O2 =0.022, balance=ar. 1271K, 15atm, = Figure 5.9 Influence of higher rate for CH 3 +HO 2 =CH 3 O+OH, new rate= 6.8 x10 13 cm 3 /mole/s, on the OH predictions by Chaos et al. [52] for n-heptane oxidation. X heptane =1000ppm, X O2 =0.022, balance=ar, =0.5, 1271K, 15atm Figure 5.10 Major MCH oxidation pathways using the Orme et al. [36] mechanism (integrated ROP), 1275 K, 16 atm, 1000 ppm MCH/O 2 /Ar, = 0.5. Refer to Orme et al. [36] for IUPAC nomenclature of species and for detailed molecular structures and pathways Figure 5.11 OH Sensitivity for MCH oxidation using the Orme et al. [36] mechanism. X MCH =1000 ppm, O 2 ( =0.5), balance=ar K, atm. (a): during ignition, (b): at early times Figure 5.12 Effect of modifications to the Orme et al. [36] mechanism predictions for MCH oxidation (combining the Libby et al. [68] recommendations (1 and 3) and the 6.788x10 13 cm 3 /mole/s value for CH 3 +HO 2 =CH 3 O+OH reaction). Initial X MCH =1000ppm, O 2 ( =0.5), balance=ar. 1262K, 15.45atm. See text for details of the Libby et al. [68] recommendations Figure 5.13 n-dodecane oxidation pathways using the You et al. [118] mechanism. X NC12H26 =1000ppm, O 2, =0.5, balance=ar, 1275K, 16atm Figure 5.14 Normalized OH sensitivity results for n-dodecane oxidation using the You et al. [118] mechanism. X NC12H26 =1000ppm, =0.5, balance=ar. 1275K, 16atm Figure 6.1 Example OH+ethylene rate measurement at 1201K, atm. Model predictions using the best-fit predictions for the rate and a factor of two variations from the measured rate are also shown Figure 6.2 OH sensitivity plot for rate measurement of OH+ethylene at 1201 K, xv

16 Figure 6.3 Example OH+propene rate measurement at 1136 K, atm. Model predictions using the best fit predictions for the rate and 50% variation from the measured rate are also shown Figure 6.4 OH sensitivity plot for rate measurement of OH+propene rate measurement at 1136 K, atm Figure 6.5 (CH 3 ) 3 -CO-OH OH+ (CH 3 ) 3 CO, k 3, rate measurement at 890 K, atm. Model predictions using the best fit predictions for the rate and 50% variation from the measured rate (k 3 ) are also shown Figure 6.6 Early-time OH sensitivity plot during OH+propene rate measurement at 890 K, atm showing the dominance of TBHP decomposition, (CH 3 ) 3 -CO-OH OH+ (CH 3 ) 3 CO, k Figure 6.7 Example low-temperature OH+propene, k 2, rate measurement at 890 K, atm. Adjustments to k 2 using the 25% uncertainty in (CH 3 ) 3 -CO-OH OH+ (CH 3 ) 3 CO are also shown Figure 6.8 OH sensitivity plot during incident shock for (CH 3 ) 3 -CO-OH OH+ (CH 3 ) 3 CO, k 3, rate measurement at 795 K, atm Figure 6.9 Example incident shock (CH 3 ) 3 -CO-OH OH+(CH 3 ) 3 CO, k 3, rate measurement at 795 K, atm. Model predictions using the best fit predictions for the rate and 50% variation from the measured rate (k 3 ) are shown. Influence on k 3 due to an assumed factor of 2 uncertainties in k 1 is also shown Figure 6.10 Arrhenius plot for OH+ethylene (k 1 ) at temperatures greater than 400K Figure 6.11 Arrhenius plot for OH+propene (k 2 ) at temperatures greater than 800 K Figure 6.12 Arrhenius plot for TBHP (CH 3 ) 3 CO+OH Figure 6.13 Arrhenius plot for OH+ethylene (k 1 ) at temperatures greater than 400K. Comparison with theory and evaluations is shown Figure 6.14 Arrhenius plot for OH+ethylene (k 1 ) at temperatures greater than 600K. Comparison with current TST calculations is shown Figure 6.15 Arrhenius plot for OH+propene (k 2 ) at temperatures greater than 650K. Comparison with theoretical calculations is shown Figure 7.1 Example OH+ 1,3-butadiene rate measurement. Initial reflected shock conditions: 1200K, 2.20 atm, 356 ppm 1,3-butadiene, 18 ppm TBHP in argon. xvi

17 Model predictions using the best fit predictions (k 1 = 5.92x10 12 cm 3 /mol/s) for the target rate coefficient and 50% variation from the measured rate are also shown. 110 Figure 7.2 OH sensitivity plot for conditions of Figure K, 2.20 atm Figure 7.3 Example OH+ 1,3-butadiene rate measurement at 1026K, 2.13 atm. Model predictions using the best fit predictions for the rate and 50% variation from the measured rate are also shown Figure 7.4 OH sensitivity plot for rate measurement of OH+1,3-butadiene at 1026K, 2.13 atm (conditions of Figure 7.3) Figure 7.5 Comparison of measured high-temperature OH+1,3-butadiene rate data with previous measurements (Liu et al. [64]) and a rate coefficient estimation (Laskin et al. [56]) Figure 7.6 Variational transition-state theory (V-TST) and measured rate coefficients in the K temperature range. Curves neglecting tunneling and variational effects are also presented Figure 8.1 OH+1-butanol rate used in latest mechanisms (Black et al. [205], Sarathy et al. [201], Moss et al. [202]) Figure A.1 Pressure time-histories from Davidson et al. [115] study near 20atm Figure D.1 3-D potential energy surfaces for the collinear reaction A+BC AB+C xvii

18 Chapter 1: Introduction 1.1 Motivation Fossil fuels account for over 80% of the primary energy consumed in the world (it is still expected to be about 70% in 2050), and nearly 60% of that amount is used in the transport sector. Hence, it is widely concluded that fossil fuels are responsible for the emission of a significant amount of pollutants in the atmosphere, including GHG (greenhouse gases). The current fuel/energy crisis magnifies the need for improving the performance of combustion engines by introducing scientific designs in which the use of chemical kinetics will be essential and critical for reducing pollutant emissions (such as, soot, CO 2 etc). There are major gaps in our knowledge of the high-temperature chemistry of real liquid carbon-based fuels. These gaps can be partially filled by reasonable theoretical chemical estimates, but benchmark experiments are needed to validate these estimations, to provide insight into chemistry which is not yet accessible by theory, and to provide highly accurate values for the most crucial reactions. One of the key emerging technologies focuses on designing an engine type, such as those based on homogeneous charge compression ignition concept (HCCI), where the chemical kinetic combustion process is controlled by temperature, pressure, and composition of the in-cylinder charge. Efforts to improve combustion efficiency and to reduce the formation of pollutants include improved engine designs, advances in emissions control technologies and the use of cleaner burning fuel formulations, all of which require detailed understanding of fuel oxidation chemistry. In this thesis, research work has focused on using state-of-the-art shock tube and laser absorption methods to investigate jet fuel and surrogate oxidation systems. Most of these studies were conducted at practical engine-relevant conditions (high pressure, low to intermediate temperatures). Some details of the relevance of these studies are given below. The development of surrogate fuels for the optimization of engine performance and the attendant chemical kinetics represents one of the important research fronts of combustion science and technology. These kinetics mechanisms accurately describe the 1

19 burning characteristics of surrogates of practical fuels such as those used in gasoline, diesel, or jet engines, and those derived from organic sources such as petroleum, biomass, coal, and/or natural gas. Studies of large realistic fuel molecules include several notable challenges in both experiment and modeling. Ignition delay time (τ ign ) is an important parameter in the combustor design of most engines, including ram-jets, scramjets, pulsed detonation engines, gas turbines, and advanced combustor concepts designed for low NOx emissions. The performance of kerosene-fueled conventional ram-jet engines has been a subject of investigation for the last 40 years, and recently, interest in the capabilities of hydrocarbon-fueled scramjet technology has increased [1,2]. In scramjets, rapid spontaneous ignition and complete reaction of fuel are required to achieve efficient combustion. Ignition characteristics also affect heat release rates, and if too rapid, can promote dynamic instabilities or choking [3]. Dependence of ignition delay time on temperature, pressure and composition is critical in describing the combustion of liquid fuels in diesel engines and combustion chambers of various types, as well as in optimizing external combustion. While ignition delay times are necessary to validate the overall behavior of reaction mechanisms, species time-histories, in particular those of important radicals such as OH, are needed to impose stronger constraints on mechanisms so that deficiencies can be identified with greater confidence. Shock tubes are nearly ideal devices for studying ignition phenomena as they provide well-controlled step changes in temperature (T) and pressure (P), well-defined time zero and τ ign, and for moderate or large diameter tubes, are generally not significantly affected by surface or transport phenomena. Volatile impurities in the shock tubes (prior to the introduction of the test gas) are also not typically a problem because of the good high vacuum characteristics and cleanliness of modern shock tubes. Laser-based diagnostic studies are non-intrusive, provide in-situ measurements (e.g., concentration of individual species including trace species, temperature, pressure and velocity), and have fast-time response (sub microsecond). The combination of shock-heating and laser detection provide a state-of-the-art test platform for combustion chemistry. Using this platform, we have conceived, defined and performed gas-phase ignition delay time, OH species time-history and rate measurement experiments. High-pressure 2

20 experiments were performed in a heated, high-pressure shock tube at Stanford (modified to deal with low-vapor pressure fuels). These studies were performed under practical engine conditions for distillate gas-phase fuels, Jet-A and JP-8, and their major surrogate components including: n-dodecane and methylcyclohexane (MCH). Ignition delay times were measured using classical shock tube methods and were combined with laser absorption measurements of species concentration time-histories. These species timehistory measurements provided important additional information about the oxidation kinetics that is not available from ignition delay times. This work was challenging, in that OH absorption experiments are difficult to conduct at high pressures due to various factors such as gas dynamic interactions and spectroscopic considerations. Detailed comparisons of these experimental data were then made using several kinetic mechanisms. The performance of the various models varied notably, even though all tested models have been validated against a variety of experimental data in other reacting configurations. This strongly illustrated the utility of and need for shock tube species time-history (and reaction rate data) for testing reaction mechanisms. Strategies were developed to improve these mechanisms and to measure some of the key OH reactions. Specifically, reactions of several stable intermediates and OH were measured and theoretical calculations were performed to support the experimental findings. 1.2 Background Jet Fuel Ignition Times Distillate fuels such as kerosenes are defined by broad composition guidelines, hence ignition delay times vary with fuel composition, among other variables, and this provides a challenge in dealing with most practical fuels. Jet fuels such as JP-4, JP-5, JP-7, Jet-A, JP-8, T-6, TS-1, and others, often contain thousands of compounds. JP-8 fuel is a military equivalent to the commercial fuel Jet-A and differs primarily by the addition of trace amounts of additives such as lubricity improvers, corrosion inhibitors, icing inhibitors, and antistatic additives [4-6]. Although the ignition characteristics of aviation fuel mixtures have been a subject of investigation for many years, there remains a critical need for experimental data on 3

21 ignition delay times of Jet-A and JP-8 fuels in air. Recent modeling efforts have been focused on developing reduced mechanisms and finding suitable surrogate mixtures for jet fuels. Challenges still remain in modeling real jet fuel behavior and in finding a suitable physical surrogate (a mixture that has generally the same physical properties as the jet fuel [6]) and a chemical surrogate (a mixture that has generally the same chemicalclass composition and average molecular weight as the jet fuel [6]) that can accurately duplicate the real fuel behavior and ignition times for a wide range of conditions. Past experimental work can be divided into two groups: flow reactors and bombs, and heated and unheated shock tubes [7,8]. Nearly all shock tube studies of kerosene/jet fuel ignition have been performed with droplets or liquid films. Ignition delay time was reported to be influenced by the following parameters: oxygen concentration, incident shock speed, droplet size and spacing, initial droplet temperature. However, droplet transport problems cause significant uncertainties in those studies. To our knowledge, there has been only one shock tube study of gas-phase jet fuel ignition. In that recent work by Dean et al. [9], ignition delay times of gas-phase Jet- A/Air mixtures were studied behind reflected shock waves in a heated shock tube. The Jet-A ignition delay time data from Dean et al. range over temperatures of 1000 to 1800 K and fall into two pressure groups: 10 and 20 atm. Two different heated shock tubes and optical arrangements were used in the study. A comparison of the ignition delay times of Dean et al. and the current study is shown later. Due to the high boiling points of large hydrocarbon components of Jet-A and JP-8 fuels, heating of the shock tube and associated mixing facilities is necessary to prepare a homogeneous gas phase mixture for ignition delay time studies, and this largely explains why so few studies have been conducted with gas-phase jet fuel using the shock tube method. This method, however, is generally preferred over other methods as it can provide reliable, reproducible measurements of gas-phase ignition delay times, amenable to a simple constant-volume modeling constraint, over a wide range of temperatures, pressures and fuel mixtures. Commercial jet fuels may contain in excess of 1000 components [6,10]. Due to the complexity of such jet fuels, it is necessary to establish simple multi-component surrogate mixtures for Jet-A, JP-8 and kerosene fuels, which can reproduce the combustion behavior of the complex fuels. The use of a simple surrogate blend, 4

22 containing a relatively small number of high-purity hydrocarbons and blended to simulate the combustion performance of a practical fuel, has the advantage of allowing fuel composition to be accurately controlled and monitored for research purposes. There is a critical need for experimental kinetic databases that can be used for the validation and refinement of jet fuel surrogate mechanisms. Shock tube studies can contribute to these databases by providing ignition delay time measurements, which describe the overall performance of mechanisms, and species concentration time-history measurements, which constrain modeling of the reaction pathways and the rates involved in these mechanisms. The jet fuel surrogate mixtures used in these detailed mechanisms approximate the chemical behavior of distillate jet fuels that contain hundreds of compounds (such as Jet-A, JP-8, and other related kerosene-based fuels), with a smaller discrete subset of compounds. The literature contains a wide variety of surrogate mixtures for jet fuels, of varying complexity, that are valid for different applications [7]. However, none of the calculations using current mechanisms and these surrogate mixtures successfully predict the combustion behavior of real fuels at all conditions. Challenges still remain in the modeling of real fuel behavior, particularly at high pressures and low temperatures, where data are needed for jet fuel as well as for surrogate mixtures and single fuel components. Based on their recent detailed review work, Dagaut and Cathonnet [8] cited the key need for high-pressure ignition measurements of kerosene-based fuels. In a very recent article, Colket et al. [11] proposed a roadmap for future development of surrogate fuels. This team of authors concluded that the main limitations in developing fundamental chemical kinetic models applicable to jet fuel surrogates have been the lack of reliable validation data sets and studies that compare real fuels and surrogate mixtures for fundamental target conditions [11] Surrogate Ignition Times Large n-alkanes constitute more than 50 % by volume of jet fuels and generally likewise, the surrogate mixtures [4]. Detailed kinetic mechanisms of these larger hydrocarbon components of jet fuels are not well understood, although the development of experimental kinetic databases for surrogate jet fuels has been a focus of much recent 5

23 research [11]. n-dodecane is widely used as the primary representative for n-alkanes in jet fuel surrogates [5], and thus there is a critical need for experimental characterization of n-dodecane oxidation. Currently even ignition delay time data are not readily available, and this has been a major factor impeding the development of an accurate chemical kinetic mechanism for n-dodecane combustion [12]. Dependence of ignition delay time on temperature, pressure and composition is critical in describing the combustion of liquid fuels in practical engines (e.g., internal combustion and gas turbine combustors) and advanced combustion chambers of various types, as well as in optimizing external combustion. Only a limited number of previous n-dodecane oxidation studies have been performed. Kadota et al. reported ignition delays of single n-dodecane droplets in quiescent air [13]. Eigenbrod et al., using a static system, observed similar ignition times for n-dodecane and kerosene droplets [14]. Segawa et al. found that the ignition delay time depended on the size of droplets and that τ ign increased monotonically with an increase in the initial n-dodecane droplet diameter [15]. Recently, Holley et al. [16] measured ignition temperatures and extinction strain rates of n-dodecane/air in a counterflow configuration under non-premixed flame conditions. We have, however, found no studies of gas-phase shock tube ignition delay time data. Cyclo-alkanes (naphthenes) are an important chemical class present not only in jet fuels, but also in other practical fuels such as gasoline and diesel, and may constitute up to 20 % by volume of jet fuels such as Jet-A/Jet-A1/JP-8, around 60% by volume of RP-1 and more than 40 % by wt. of diesel fuels [4,6,17,18]. Cyclo-alkanes in diesel fuels have been found to influence particulate matter (PM) emissions [18], and the soot precursor production potential of cyclo-alkanes (via formation of polycyclic aromatics) is much higher than that of normal- and iso-paraffin compounds [19]. For many years, the combustion and kinetic behavior of straight chain and branched alkanes has received significant attention, while cyclo-alkanes, on the other hand, have received, until recently, only scant attention [20]. Methylcyclohexane (MCH) is one of the simplest cyclo-alkanes and is widely used to represent the cyclo-alkane fraction of jet fuel surrogates [5,21-24]. It has also been proposed as a fuel for scramjets, because in the presence of a catalyst MCH can be endothermically dehydrogenated to form toluene and hydrogen, thus providing a 6

24 significant heat sink (~2190 kj/kg) for cooling the engine [25]. Very few studies concerning MCH combustion exist which can be utilized for surrogate fuel development, though recently there have been a variety of experimental and kinetic modeling studies conducted for other types of cyclo-alkanes [20,26]. The kinetics of cyclo-alkanes is regarded as critical to properly representing the reactivity of gasoline and kerosene-based fuels [27]. Previous researchers studied some aspects of the kinetics of MCH in a variety of different experimental facilities, such as, low-temperature oxidation in a flow tube [28], unimolecular decomposition in a pyrolysis apparatus [29], oxidation in a single-cylinder production-type engine [30], high-temperature ( K) pyrolysis and oxidation in a turbulent flow reactor [31], oxidation in a pressurized flow reactor [21], ignition delay times ( ign ) in rapid compression machines (RCM) [22,32,33], decomposition in laminar diffusion flames [34], and ignition in shock tubes [35,36]. Granata et al. [37] reported a semi-detailed model of MCH pyrolysis (validated using the measurements of Zeppieri et al. [31]) and included cyclo-alkanes as reference components to broaden their surrogate model s capabilities for heavy practical fuels such as jet fuels, kerosene, and diesel oils. Granata et al. [37] pointed out that their modeling study was specifically constrained by the lack of experimental data on the oxidation of MCH. High-pressure (above 4 atm) shock tube ignition delays of MCH have not been reported so far High-pressure OH Time-histories During Surrogate Oxidation Species time-history measurements of OH would provide strong kinetic targets for the validation and refinement of detailed models and will lead to more accurate ignition time predictions, which are intimately tied to the actual radical pool kinetics. Hydrocarbon ignition is, to a large extent, controlled by the chemistry of the small transient radical pool (H, OH, CH 3, etc.), and in particular, very few or no data are available for these species in shock tube studies of relevant surrogate components. At higher pressures, most current mechanisms have been validated only against the measured yields of the more stable intermediates, and hence the importance and role of the small radical pool has not been tested. As well, at higher temperatures, where the initial removal of hydrocarbons is 7

25 strongly affected by decomposition as well as H-atom abstraction reactions, species timehistory measurements of OH can provide information about both these pathways. In the current study, we have measured OH concentrations during MCH, n-dodecane and n- heptane oxidation near 15 atm in argon. n-heptane is one of the most widely researched higher-order hydrocarbons, mainly because of its application as a primary reference fuel, as a homogeneous charge compression ignition (HCCI)-relevant fuel, and as a diesel fuel. Numerous experimental and modeling studies of high-pressure n-heptane oxidation and pyrolysis can be found in the literature [20,38-40]. Starting with the early kinetic model of n-heptane combustion developed by Coats and Williams [41], well-established kinetic reaction mechanisms (some of which will be examined later in this thesis) for this fuel are widely available. However, the only OH concentration measurements in n-heptane/o 2 /Ar systems were conducted by Davidson et al. [42] for pressures between atm and temperatures from 1540 to 1784 K in a low-pressure shock tube at our laboratory. Davidson et al. [42] examined some of the n-heptane mechanisms and noted that mechanisms were only moderately successful in duplicating the OH time-histories. By comparing with the ethylene yield measured (using a microwave-lamp absorption diagnostic at 174nm) in the shock tube studies of n-heptane pyrolysis by Horning et al. [43] at atmospheric conditions, Davidson et al. [42] noted that the predicted ignition times in n-heptane oxidation appeared to follow the predicted time scale of ethylene removal in n-heptane pyrolysis systems. This suggested that the OH time histories are strongly linked to ethylene populations, and Davidson et al. [42] found that the ability of the n-heptane mechanisms to model OH is directly linked to their ability to model ethylene concentrations. Hence, accurate OH measurements can be used to place constraints on the initial fuel decomposition processes, the size of the radical pool that exists during the induction regime before ignition, and the reaction rates of major product species such as ethylene. Nonetheless, there exist no OH measurements for high-pressure n-heptane oxidation systems. Various researchers from our laboratory have measured OH time-histories during fuel oxidation for different hydrocarbons, however, OH absorption experiments are difficult to conduct at high-pressures due to various factors such as gas dynamic interactions and 8

26 spectroscopic considerations. Consequently, most of those studies, with the exception of Petersen et al. [44] (in methane), were carried out near atmospheric pressures or at pressures less than 10 atm using low-pressure shock tubes. Current data are the first OH species time-history measurements in MCH and n-dodecane oxidation and the first OH species time-history measurements in high-pressure n-heptane oxidation systems Reaction of OH with Alkenes Simplest alkenes such as, ethylene (C 2 H 4 ) and propene (C 3 H 6 ), are the most important intermediates in the oxidation of hydrocarbons fuels [45] and are formed in large quantities during combustion of hydrocarbons from methane to real fuels in all practical engines. Also, ethylene and propene are present in practical fuels and are also emitted into the atmosphere through anthropogenic and natural sources. Alkane fuels rapidly decompose to simple olefins (before ignition) during high-temperature oxidation [45]. The oxidation kinetics of ethylene and propene are very important to the hierarchical development of the kinetic mechanisms of real fuels [45-48]. Alkenes contribute to soot production (and other pollutant formation) and therefore strategies for mitigating pollutant formation in advanced combustion systems depend on alkenes oxidation chemistry. Knowledge of alkene oxidation reactions is critical to the development of accurate modeling of combustion processes such as ignition, heat release, etc., at high temperatures. Accurate determinations of ignition times and species time-histories for hydrocarbon fuels are strongly sensitive to olefin concentration levels (mostly ethylene and propene) and oxidation rates [45]. Hydroxyl (OH) radical is an important transient reactive radical during combustion and in atmospheric chemistry, and it is also widely accepted that OH radical oxidation of ethylene and propene is the major oxidation route for these molecules under atmospheric and combustion conditions [45,49,50]. Ethylene and propene react with OH to form various products including water (the final product): C 2 H 4 +OH Products, and C 3 H 6 +OH Products. Because it is the most important (and the simplest alkene) stable intermediate during combustion of higher hydrocarbons, extensive experimental and theoretical studies have been conducted on OH+ethylene. Table 1.1 lists all studies above 500 K along with some 9

27 of the low temperature studies of OH+ethylene. Despite these measurements, relatively large variations in the rates (magnitude and activation energy) for C 2 H 4 +OH exist in recent combustion kinetic mechanisms for practical fuels (see Figure 1.1). Shown are rates used in recent hydrocarbon mechanisms such as, JetSurF 1.0 [46], Westbrook et al. [47], Ranzi [48], Galway natural gas mechanism III [51], and Chaos and Dryer [52]. We believe this lack of consensus among modeling community arises from the scatter in the experimental studies listed in Table 1.1. Relatively very few studies have been conducted on the reaction of OH with propene (see Table 1.1). To the best of our knowledge, there has been no experimental study of OH+propene above 1210 K. Hence investigations at higher temperatures appear to be warranted for both OH+ethylene and OH+propene. 1E K 1000 K 667 K 500 K k 1 [cc/mol/s] 1E12 1E11 1) JetSurF 1.0, ) Westbrook et al., ) Natural gas III (Galway), ) Ranzi, ) Chaos et al., E K/T Figure 1.1 Comparison of rate constants for C 2 H 4 +OH used in several current mechanisms. Variations of a factor of ~5.5 are evident at high temperatures near 1000 K. Shown are rates used in JetSurF 1.0 [46], Westbrook et al. [47], Ranzi [48], Galway natural gas mechanism III [51], and Chaos et al. [52] OH+1,3-Butadiene 1,3-Butadiene (1,3-C 4 H 6 ) is an important stable intermediate formed during the combustion of hydrocarbons, from methane to real fuels, in practical engines. It is also a hazardous, carcinogenic, toxic pollutant and genotoxic (in humans and other mammals) chemical, which is widely used in petroleum and rubber industries, and emitted into the

28 atmosphere from sources including tobacco smoke, forest fires, automobile exhaust, and gasoline evaporative emissions [53-55]. Because of this, there is a need to include the oxidation kinetics of 1,3-butadiene in the hierarchical development of the kinetic mechanisms of real fuels [26,31,36,45,56]. A better knowledge of the oxidation kinetics of 1,3-butadiene (as the simplest conjugate olefin) would help improve the understanding of the role of complex olefins in combustion, soot formation and toxic emissions [57,58]. There have been many studies of the most important of the 1,3-butadiene oxidation reactions, OH+1,3-C 4 H 6 =products. But these have been limited mainly to low temperatures [53-55,59-67]. There also have been several high-temperature studies of 1,3-butadiene oxidation and pyrolysis in flames, flow reactors, and shock tubes [56,57,68,69], but these studies have not directly addressed determination of the rate coefficient of OH+1,3-butadiene. The only experimental study (> 500K) for the overall rate of this reaction we have found was conducted in a flow system using pulsed radiolysis coupled with microwave resonance absorption to monitor the OH concentration in the range K [64]. To the best of our knowledge, there has been neither an experimental nor a theoretical study of this reaction above 1203 K. 1.3 Scope and Organization of Thesis The primary objective of this work was to conduct high-quality experiments using state-of-the-art shock tube and laser absorption methods to investigate jet fuel and surrogate oxidation systems. Ignition times and species time-histories were measured under engine-relevant conditions and low-uncertainty measurements of the reactions of OH with several stable intermediates were carried out. Current data were used as validation targets and improvements to several surrogate mechanisms were provided using detailed kinetic analysis. Theoretical calculations of the measured rates were provided which supports our experimental findings. Chapter 2 describes the experimental apparatus, OH laser diagnostics and methods used to model the shock tube data. Chapter 3 presents ignition delay time measurements in jet fuels (Jet-A/air and JP-8/air) at high pressures using the heated HPST. Predictions of several jet fuel kinetic mechanisms using surrogate mixtures were analyzed. Chapter 4 contains high-pressure ignition delay time measurements in two n-dodecane/air and 11

29 MCH/air, which are the two most important single-component jet fuel surrogates. Available kinetic mechanisms for these two fuels were analyzed for their predictive capabilities. Chapter 5 describes high-pressure OH time-histories measured during oxidation of three single-component surrogates (n-dodecane, MCH, and n-heptane) diluted in argon using heated HPST. Detailed comparisons of experimental data with predictions of available kinetic mechanisms were made and procedure was shown to improve their predictive qualities. Chapter 6 presents measurements of the reactions of OH radicals with two important alkenes (ethylene and propene) behind reflected shock waves. Also discussed in this chapter are canonical transition state theory (TST) calculations for these reactions. Chapter 7 describes measurements of the reaction of OH with 1,3-butadiene. Theoretical results of the rate coefficient and the branching fractions for the H-abstraction channels of the target reaction were also presented. Chapter 8 summarizes the thesis and suggests future work. Appendix section details modes of ignition, shock tube nonideal effects, and TST theory. 12

30 Table 1.1 High-temperature OH+ethylene and OH+propene studies in literature. Reference Temperature Range Experimental Method and Comments OH+ethylene Bott and Cohen [50] 1197 K Reflected Shock Tube (S.T.), 1.04atm, OH resonance microwave absorption at 309nm Smith [70] 1220 K Laser pyrolysis/laser fluorescence Bradley et al. [71] ~1300 K Incident S.T., OH UV Lamp, rate was determined relative to methane + OH Baldwin et al. [72] 813 K Reaction vessel, pressure and gas sample analysis using method similar to gas chromatograph Bhargava and Westmoreland [73] K Laminar flames, molecular-beam mass spectrometry (MBMS) Westenberg and Fristrom K Flames, electron spin resonance (ESR) spectroscopy [74] Tully [75] K Flash photolysis, laser induced fluorescence (LIF) Tully [76] K Laser photolysis, LIF Liu et al. [64,77] K Pulse radiolysis/ OH resonance microwave absorption, 1atm Westbrook et al. [78] K Jet-stirred reactor (JSR) Greiner [79] K Flash photolysis/ kinetic spectrograph Fulle et al. [80] K Laser flash photolysis/ saturated LIF Srinivasan et al. [81] K Reflected S.T., OH UV Lamp Baldwin et al. [82] 773 K Reaction vessel, pressure and gas sample analysis using method similar to gas chromatograph (GC) Diau and Lee [83] K Laser photolysis/ LIF Zellner and Lorenz [84] K Laser photolysis/resonance fluorescence Hoare and Patel [85] K Reaction vessel/gc Avramenko and Lorentso [86] OH+propene K Discharge flow system. OH generated from H 2 O vapor. OH measured by U.V. absorption. Source of OH at fault Bott and Cohen [50] 1204 K Reflected S.T., 1.04atm, OH resonance microwave absorption at 309nm Tully and Goldsmith [87] K Laser photolysis, LIF Smith et al. [88] K Laser pyrolysis, LIF Baldwin et al. [89] 773 K Reaction vessel, pressure and gas sample analysis using method similar to gas chromatograph. Relative rate method using OH+tetramethylbutane reaction Atkinson and Pitts Jr. [90] K Flash photolysis, LIF Yetter and Dryer [91] 1020 K Flow reactor (1atm), GC 13

31 Chapter 2: Method This chapter describes the shock tubes and the OH laser diagnostics used in this study. 2.1 Shock Tube Facilities High-pressure Shock Tube All high-pressure (above 10 atm) ignition delay times and OH time-histories were measured behind reflected shocks using Stanford s heated, high-pressure shock tube (HPST), which was designed to attain pressures as high as 600 atm. The shock tube driver section is 3 m long with a 7.5 cm internal diameter. Helium was used as the driver gas, except for the low-temperature cases (< 900 K) where a tailored driver gas mixture (20-30% N 2 in He) was used. Diaphragms were made of aluminum of 1.25 to 2 mm thickness with cross-scribing. The stainless steel driven section of the HPST is 5 m long with a 5 cm internal diameter and was wrapped with thin copper sheets and heated to 100 C using 13 separate heating zones to produce uniform temperature along its length; this prevents condensation of the fuel/air mixture. Uniform temperature (±3 ºC) along the length of the shock tube is achieved through a series of heaters controlled using Watlow controls. Before introduction of the test mixture, ultimate pressures of less than 10-5 torr and leak-outgasing rates of less than 10-4 torr/min are regularly achieved using Varian TM SD 250 roughing pumps and a turbo-molecular pump (Varian TM Turbo-V 250). End-wall movements during experiments were limited to less than 1 mm by attaching the last section of the driven tube to a 2 ton concrete and steel mass. A fuller description of the shock tube can be found in Petersen [92]. The incident shock speeds were determined using six fast-response piezo-electric pressure transducers (PCB 113A), spaced at approximately 30 cm intervals over the last 2 meters of the shock tube, and five time-interval counters (Phillips PM6666). Uncertainty in the time interval measurement is ±0.3 μs. The velocity of the incident shock at the end wall is then determined by extrapolation. Typical shock attenuation rates, defined as the normalized slope of axial velocity extrapolated to the end wall (in %/meter), ranged from 0.2 to 2 %/m for the current experiments. Fill pressure (P 1 ) is monitored using two static 14

32 pressure transducers (Setra Model 280E). A K-type thermocouple protruding slightly from the sidewall midway along the driven section was used to determine pre-shock test mixture temperature (T 1 ). Using P 1 and T 1, conditions behind the reflected shock wave (temperature T 5, pressure P 5 ) were determined by the one-dimensional normal shock equations and the Sandia thermodynamic database of Kee et al. [93] including additional thermodynamic properties information on liquid Jet-A, OH and surrogates as recommended by Burcat and Ruscic [94]. Ignition pressure was monitored using a piezoelectric pressure transducer (Kistler model 603B1) located 10 mm from the end-wall. For ignition experiments, the emission from OH* (using a Schott Glass UG-5 filter with >95% transmission at 306 nm), detected using a lens/slit setup with a Thorlabs PDA55 detector, was monitored at an observation window located at the same axial location. The axial spatial resolution of the detector system (<5 mm) was determined from the optical design calculations of Flower [95]. A schematic of the emission set-up is given in Figure 2.1. For the current reflected shock temperatures and pressures, real gas corrections to the temperatures were insignificant and within the experimental uncertainty for the HPST [96]. Vibrational relaxation in air was also taken into consideration, when computing the reflected shock temperatures, since the vibrational relaxation state of O 2 and N 2 was found to have significant effect on the predicted temperatures and pressures using the shock code. Temperatures and pressures calculated in this study assume full vibrational relaxation of the shocked gases in both incident and reflected regime. The experimentally measured pressures were found to be consistent with the predicted pressures when full vibrational relaxation was assumed, as was also seen by Gauthier et al. [39]. Care was taken to minimize the uncertainties caused by reflected shock bifurcation on time-zero definition, following the recommendations from Petersen [92]. Strong ignition was observed in all ignition time experiments conducted owing to the large fuel concentrations employed. As a consequence, the combustion wave may transition to a detonation wave, thereby affecting the ignition delay time measurement further down the side wall by shortening the distance and arrival times between the reflected shock wave and the combustion front. However, for the present diagnostics located at 10 mm from the endwall, and for the reflected shock temperatures and pressures studied, the estimated 15

33 discrepancy was less than 10 μs for all cases, and was simply included in the uncertainty analysis Low-pressure Shock Tube All reaction rate experiments were performed behind reflected shock waves in a stainless steel, high-purity, helium-driven, unheated low-pressure shock tube (LPST) with inner diameter of cm. Lexan polymer diaphragms were used and the shock tube test section and mixing manifold were routinely evacuated below 1 Torr using two turbomolecular pumps. The shock tube leak-plus-outgassing rate was less than 5 Torr/min. Incident shock velocities were measured using five piezoelectric transducers (PCB 113A) spaced axially along the last meter of the tube and linearly extrapolated to the endwall. Average incident shock wave attenuation rates were between 0.6% and 2% per meter. Ideal shock relations and the thermodynamic database from Burcat and Ruscic [94] were used to calculate reflected shock temperature and pressure (T 5 and P 5 ). Uncertainties in the calculation of the initial reflected-shock temperature and pressure were typically less than 0.7% and 1%, respectively, and arose primarily from the 0.3% uncertainty in the incident shock velocity determination. This shock tube has been extensively used for reaction rate measurements in our lab (further details can be found in [97-99]). 2.2 OH Laser Absorption Diagnostics Absorption spectroscopy of the OH radical near nm is well-established at combustion temperatures and near-atmospheric pressures. The OH absorption measurements, at the same axial sidewall location where P 5 was measured, were made using narrow-linewidth ring-dye laser absorption close to the R-branchhead (overlap of the R 1 (8), R 1 (9) and R 1 (10) lines at high-pressures) near nm (32630 cm -1 ) of the OH A 2 Σ+-X 2 Π (0,0) system in high-pressure experiments. This particular peak wavelength was selected as it has the strongest high-pressure OH absorption feature at high pressures. In low-pressure experiments (1-10 atm), OH radicals were monitored using the R 1 (5) absorption line near nm as it has the highest peak and was well isolated from other peaks. Visible light at nm (25-30 mw) was generated in a Spectra-Physics 380A ring-dye laser cavity by pumping Rhodamine 6G dye with a 5 W, 16

34 532 nm, cw beam produced by a Coherent Verdi laser. The visible beam was intra-cavity frequency-doubled using a temperature-tuned AD*A crystal, generating 1-2 mw of UV light. A small part of the visible output from the laser (the laser has an instantaneous linewidth of a few MHz) at the fundamental wavelength was used for peaking the laser power as well as for monitoring laser mode quality in a scanning interferometer and reading the absolute laser wavelength using a Burleigh WA-1000 wavemeter. Highpressure absorption measurements require attention to two phenomena not generally important at lower pressures: stress-induced anisotropy within the observation windows, and laser propagation through flow-induced perturbations. The current high-pressure optical setup incorporated recommendations from Petersen [92] to deal with these effects (choice of the optical components and their relative positioning can be optimized to minimize these effects). The low-pressure optical setup schematic was same that used in [99]. Briefly, UV light was separated into two beams: a reference beam I o, and a transmitted beam I, which passed through the sidewall measurement location, which was at 10 mm and 20 mm from endwall, for HPST and LPST, respectively. Two sapphire windows provided optical access to the test section for the UV absorption experiments. Windows were pre-stressed in a separate rig to reduce stress-induced anisotropy. Common-mode rejection of laser intensity fluctuations was performed by balancing the two beams prior to each run using neutral density filters. Thorlabs PDA 55 UV detectors, specially modified to accommodate large area Hamamatsu S detectors (with an effective active area of 13.2 mm 2 ), monitored both the UV beams. A spectral filter (Newport FSR- UG11, Schott Glass, UV Bandpass) was used with the detector to reduce interference from broadband emission during ignition. All data were recorded using a high-resolution data acquisition system at a sampling rate of 2MHz. The lens/detector setup was optimized to minimize flow-induced perturbations (beam steering) by insuring that the beam waist in the shock tube was positioned away from the boundary layer. Additional details of the two-beam method and the OH ring dye laser absorption diagnostic may be found in [92,97,99]. 17

35 Computer Data Acquisition HPST L I PDA 55* IR ND PDA 55* BS I 0 L IR L M UV M 532 nm L nm Coherent Verdi Pump laser S-P 380 Tunable ringdye laser Power meter Visible BS nm Si Detector Scanning etalon BS Wavemeter M IR Absorption Set-up Laser source Laser beam quality L = lens; ND = neutral density filter; BS = beam splitter; IR = iris; M = mirror; PDA 55* = UV enhanced photodiode; I = diagnostic beam; Io = reference beam Emission Set-up PDA 55 Kistler PZT L Slit Computer Data Acquisition SHOCK TUBE Figure 2.1 Experimental set-up for HPST experiments. OH mole fractions (X OH ) were determined from the measured absorption using the Beer-Lambert relation: I/I o =exp{-k ν P 5 X OH L), which assumes attenuation of incident radiation by a non-saturating, linearly-absorbing medium. P 5 (atm) is the total reflected shock pressure and L is the path length (5 cm and cm, for HPST and LPST, respectively). The well-characterized absorption coefficient, k ν, was calculated incorporating the collision-broadening and collision-shift parameters measured by Herbon [97] for the R 1 (5) line and by Davidson et al. [100] at high-pressures for R- branchhead. Laser-off measurements were also taken to confirm the absence of any significant contribution of emission to the absorption signal for the range of current 18

36 experiments. Consequently, off-line and emission corrections were not needed or applied. Several uncertainties contribute to the determination of X OH with uncertainties in temperature (T 5 ) having the largest effect. The other main uncertainty comes from uncertainty in determining the laser wavelength. However, k ν values are relatively constant for small wavelength fluctuations at current conditions. The overall estimated uncertainties in the inferred values of X OH are ±5% and ±3% at high- and low- pressures, respectively. 2.3 Modeling Shock Tube Data In comparing modeled and experimental ignition, OH time-history and rate coefficient data, the model calculations are done for homogeneous, adiabatic conditions behind reflected shock waves, with a constant-volume constraint using CHEMKIN [101]. Shock tubes produce homogeneous mixtures, and behave like near-ideal constant-volume reactors up to the time of ignition. Hence, the constant-volume, the constant-internalenergy constraint (constant U,V) is a good assumption for the purpose of ignition delay time calculations (typically less than 1-2 ms); details about the implications of this assumption for the later stages of ignition process are described in Davidson and Hanson [102]. However, at long test times, even in the absence of reaction, the reflected shock pressure (P 5 ) typically increases slowly (approximately linearly with time in our facility.) This rise is caused by non-ideal effects such as incident shock attenuation, boundary layer growth, and interaction of the reflected shock wave with the side-wall boundary layer (see Petersen [92] for details about the non-idealities in HPST). We developed a model in our lab, CHEMSHOCK [103], which combines CHEMKIN with an isentropic compression of the test gas mixture consistent with the actual pressure measured during the experiment. The influence of these non-ideal effects (if present) on the measured/modeled ignition data can be estimated using the CHEMSHOCK model. 19

37 Chapter 3: High-pressure Jet Fuel Ignition Times This chapter describes ignition delay time measurements in Jet-A/air and JP-8/air at high pressures using the heated HPST. Initial reflected shock conditions were as follows: temperatures of K, pressures of atm, equivalence ratios of 0.5 and 1, and oxygen concentrations of 10 and 21 % in synthetic air. As a key step toward defining or validating a surrogate composition for jet fuels, comparisons with the predictions of several current kinetic mechanisms are also made. 3.1 Jet Fuel/Air Mixture Preparation Jet fuel can vary widely in its chemical properties. A recent gas chromatograph results (Dean et al. [9]) for a Jet-A sample provided by U.S. Oil and Refining Co. identified 167 compounds and suggests an approximate formula C H Many heavy components were not recognized in the study. In another study by Gueret et al. [104], the general formula, obtained using a mass-detector after separation on a capillary column, was found to be C 11 H 22. Throughout this study we have assumed the widely specified chemical formula of C 11 H 21 and a density of 0.81 gm/cm 3 for all the jet fuels tested [6,9]. The Jet-A Composite Blend used in this study (#04POSF4658) was analyzed by Shafer et al. [105] using the ASTM D2425 GC-MS [106] technique and had the following main components by volume: paraffins (normal + iso) 55.2%; monocycloparaffins 17.2%; dicycloparaffins 7.8%; alkyl benzenes 12.7%; indans+tetralins 4.9%; substituted naphthalenes 1.3%. The JP-8 fuel used in this study had the following approximate characteristics: cetane number of 43.3, density of 0.8 kg/l, hydrogen content of 13.9 (wt. %), aromatics composition of 13.9 (vol %), olefins composition of 0.5 (vol %), and saturates composition of 84.2 (vol %). It is important to vaporize the jet fuels as completely as possible to reduce any errors in the fuel-air mixture calculations. A new liquid fuel mixing facility was designed for this purpose and was attached to the shock tube. Fuel/air mixtures were prepared in a 12.8 liter, magnetically-stirred, thermally insulated, stainless-steel mixing tank. For all 20

38 experiments, the mixing tank and connecting gas lines to shock tube were heated (using a custom-made heating blanket provided by HTS/Amptek) to avoid condensation of the fuel and to achieve higher fuel concentrations. The mixing tank was well-insulated to avoid any spatial temperature non-uniformity, thereby avoiding surface temperature variation and local condensation points inside the tank. In all the current experiments, quantities of liquid Jet-A (JET-A Composite Blend #04POSF4658 supplied by T. Edwards, AFRL-WP) and JP-8 (supplied by P. Schihl, ARL) were measured and added volumetrically into the mixing tank using a gas-tight syringe (Hamilton 1010TLL) and allowed to evaporate until a steady pressure (monitored by a MKS 690A Baratron pressure transducer) was achieved. For the entire study, air refers to synthetic (dry) air consisting of 21 % O 2 and 79 % N 2 (i.e., X N2 =3.76 X O2 ). Research grade synthetic air (Praxair N 2, O 2, ) was added slowly in accordance with a procedure outlined by Horning et al. [43] to inhibit condensation of fuel that could lead to inaccuracies in reported fuel composition. The mixture was stirred using a magnetically driven vane assembly, typically for 3 hours, before the first shock wave experiment. Before use in this study, the heated mixing facility was characterized using two single component fuels, iso-octane and n-heptane, whose vapor pressure are known accurately (close to 35 torr for both fuels at 20 C, as given in Lide [107]). For iso-octane and n- heptane, a fixed volume of the liquid fuel (up to 10 cc) was added to the mixing tank at 85 C. The measured pressure using the Baratron, and the ideal-gas predicted value (assuming complete vaporization) were found to be in very close agreement for these fuels at this temperature (implying no significant wall adsorption or incomplete vaporization or other loss mechanisms). For both Jet-A and JP-8, the vapor pressure is very low (less than 2 torr at 20 C), when compared to the vapor pressures for iso-octane and n-heptane. With the objective of finding the temperature to which the mixing tank and fuel lines should be heated so as to avoid any condensation of the jet fuels, 1 cc of the jet fuel was added to the tank at different temperatures. Comparison of the measured and the predicted (assuming complete vaporization) pressures showed that the mixing tank and the connecting lines had to be heated to 125 C to ensure that there was no further impact on fuel vaporization. A discussion about the effect of different mixing times for 21

39 the fuel/air mixtures inside the tank and different shock tube temperatures on the ignition delay time measurements is provided later. 3.2 Measuring Ignition Delay Times The ignition delay time is defined in this study as the time interval between the arrival of the reflected shockwave and the onset of ignition at the sidewall observation location. The arrival of the reflected shockwave was determined by the step rise in pressure, and the onset of ignition was determined by monitoring both the pressure history and the emitted light corresponding to OH* emission. The onset of ignition from the pressure history and OH* emission were defined by locating the time of steepest rise and linearly extrapolating back in time to the pre-ignition baseline (for details regarding emission measurements, see Hall et al.[108]). The two methods give very similar results (± 3%) and ignition delay times are readily identifiable with both diagnostics; example data are shown in Figure 3.1. The overall uncertainty in the ignition delay time measurements was 15%, which was dominated primarily by uncertainties in the temperature (less than 1.8 %, and due mainly to uncertainties in fuel composition and in the endwall velocity). Larger uncertainties may exist in the post-shock temperatures of the tailored gas mixtures, and are discussed in a later section. 80 JP-8/air, = K, 22.2 atm Pressure [atm] ign =615 s Pressure 0 OH* Emission Time [ s] Figure 3.1 Example JP-8/air ignition delay time data (driver gas: He). Initial reflected shock conditions, 1019 K, 22.2 atm, = 1. 22

40 3.3 Jet Fuel Ignition: Results and Discussion The results for ignition delay times versus inverse temperature are discussed in this section and are presented in the following order: comparison with previous measurements; comparison of different fuel types (Jet-A and JP-8); pressure scaling; surrogate modeling prediction; effect of variation of equivalence ratio and oxygen concentration; and lowtemperature (NTC region) ignition delay time data. All current experimental results are summarized in Table 3.1 and Table Comparison with Previous Measurements The ignition delay time data for Jet-A/air (Jet-A Composite Blend #04POSF4658) over a range of temperatures for reflected shock pressures (P 5 ) from 18 to 36 atm have been normalized to 20 atm assuming τ ign ~ 1/P and are presented Figure 3.2 (solid circles). These measurements show relatively good agreement with the one earlier (Dean et al. [9], atm) shock tube study for gaseous Jet-A (open circles). Over the temperature and pressure range (10-36 atm) of the current and the Dean et al. [9] data, a normalizing pressure scaling of 1/P correlates the data very well. However, it must be noted that different sources of jet fuel were used in current experiments and in those of Dean et al. [9]; see earlier section for additional fuel information. Hence, to perform a comparative study of τ ign measurements from different facilities for same Jet-A fuel, we also tested the Jet-A fuel used by Dean et al. [9]; results are presented as open square symbols in Figure 3.2. The HPST was heated to 80 C (same temperature as in the Dean et al. [9] study) for these experiments. The data generated in the current study and those recorded by Dean et al. [9] are in very good agreement for this Jet-A fuel, which in turn are in good agreement with the ignition delay time data for Jet-A Composite Blend #04POSF4658. Hence, it could be concluded that under the current experimental conditions, Jet-A ignition delay times do not differ between these two Jet-A fuel sources. For all the subsequent Jet-A/air discussions in the current chapter, only data for the Jet-A Composite Blend #04POSF4658 is presented. 23

41 1666 K 833 K Ignition Delay Time [ s] Jet-A/air, =1.0 Data scaled to 20 atm using P /T [1/K] Figure 3.2 Ignition delay times including previous data for Jet-A/air for phi=1. Solid circles: current work for Jet-A POSF4658; open circles: Dean et al. [9]; open squares: current work for Dean et al. [9] Jet-A fuel. As a further point of comparison, there does not appear to be a significant difference in the ignition delay time measurements when the shock tube was heated to T 1 = 80 C or 150 C (both were employed in Dean et al. [9]) or T 1 =100 C (current work). This similarity may be attributed to the fact that the ignition process is not strongly affected by the presence or absence of small amounts of heavier (~C 20 ) fuel components that may be volatilized (T 1 =100 C and T 1 =150 C) or may not be totally volatilized (T 1 =80 C) in the experiments. Larger non-volatized hydrocarbons are present only in trace amounts, and may not be a factor in determining the overall ignition process. The possibility of fuel cracking from long mixing times at elevated temperatures is a concern in kinetics studies of jet fuels (Holley et al. [109]). However, in the current experiments, no measurable change in ignition delay time data was observed for fuel/air mixing times between 2.5 to 8 hours. 24

42 3.3.2 Comparison of Different Fuel Types A comparison of the measured ignition delay times for Jet-A/Air and JP-8/Air mixtures are shown in Figure 3.3 for an equivalence ratio of 1.0. The low scatter of the data shown in Figure 3.3 enables discernment of small differences in ignition delay times in fuel types. There is very good agreement between ignition delay times for both fuels at the high and low temperatures. JP-8 ignition delay times appear to be approximately 10% shorter than those of Jet-A for temperatures near 1000K. However, owing to our experimental uncertainties, caution must be applied when utilizing this finding. Since jet fuels meet only a very broad range of fuel specifications, this apparent difference in the ignition delay time behavior with respect to fuel type or source is very hard to quantify. The combustion characteristics such as extinction and autoignition of Jet-A and JP-8, based on measurements in laminar non-premixed flows, were found to be very similar [23]. But, [110] found that variations in cetane index, aromatic and naphthene content, and fuel additives do impact auto-ignition behavior in a single-cylinder engine. In summary, these comparison experiments suggest that the combustion performance of jet engines could be affected by variations in the fuel source or type, and more detailed studies of tightly controlled fuel and surrogate mixtures are clearly needed. Ignition Delay Time [ s] 1176 K Data scaled to 20 atm using P =1.0 Jet-A/air JP-8/air 870 K /T [1/K] Figure 3.3 Ignition delay times for Jet-A/air and JP-8/air, phi=1. 25

43 3.3.3 Pressure Scaling When experimental conditions vary over a wide range of pressures, assuming a powerlaw dependence to the pressure scaling leads to a more uniform graphic presentation of the ignition delay time data. The pressure dependence of ignition delay times in Jet-A/air mixtures is shown in Figure 3.4. A simple 1/P dependence was found for ignition delay times from 850K to 1250K for the pressure range atm based on the scaling of the data points. This 1/P dependence has been observed in many previous studies with different experimental facilities [7]. A regression analysis of all the current data for Jet- A/air and JP-8/air is also shown, which shows τ ign ~ P -0.98, which is very close to the 1/P behavior cited previously. However, in all the results presented in the current chapter a simple 1/P dependence to scale ignition delay times was used. In the current data, there is slight evidence of negative temperature coefficient (NTC) type ignition delay time rolloff below 1000K. NTC behavior is attributed to the formation of peroxy radicals, and large n-alkanes present in the jet fuels have a stronger tendency to show NTC behavior than small or branched alkanes [111]. Stronger pressure dependence is expected in the NTC region K 833 K 1250K K 22 atm Ignition Delay Time [ s] atm 50 atm Jet-A/air = 1.0 Ignition Delay Time [ s] All data scaled to 20 atm using P /T [1/K] /T [1/K] Figure 3.4 Pressure scaling of (phi=1) τ ign : left- Jet-A/air; right- Jet-A/air and JP-8/air. 26

44 3.3.4 Jet Fuel Surrogate Modeling Prediction The current data set provides critical validation targets for jet fuel kinetic mechanisms and surrogate models. Predictions of the following four available kerosene-based kinetic mechanisms are analyzed: 1) Ranzi [48], Politecnico di Milano, Italy, JP-8, 280 species, 7800 rxns; 2) Zhang et al. [19], University of Utah, USA, JP-8, 208 species, 1087 rxns; 3) Dagaut and Cathonnet [8], CNRS, France, Kerosene, 209 species, 1673 rxns; 4) Lindstedt and Maurice [112], Imperial College, UK, Kerosene, 154 species, 947 rxns. The surrogate mixtures suggested by [8] and [112] were used to model the ignition delay times with their respective mechanisms (see Table 3.3). The Ranzi [48] and the Zhang et al. [19] mechanisms did not specify a particular surrogate mixture. Hence, the most widely accepted JP-8 surrogate mixtures from Violi et al. [5] were used to model the ignition delay times predictions with the Ranzi [48] and the Zhang et al. [19] mechanisms (Table 3.4 To further investigate the influence of various representative components of jet fuel compositions on the ignition delay times, two modified Violi #3 mixtures were used: Stanford A, with less n-alkanes and more branched alkanes compared to Violi #3; and Stanford B, with less n-alkanes and more aromatic components compared to Violi #3. Figure 3.5 presents the modeling results for ignition delay time using all the mechanisms for Jet-A/air at 20 atm and Ф=1.0. In Figure 3.5, the Violi #3 surrogate was used with the Ranzi and the Zhang et al. mechanisms. The ignition process in heavy fuels is a multi-step process, starting with the rapid decomposition of the fuel followed by the slow decomposition of the intermediate products and the exponential growth of the radical pool. The Lindstedt and Maurice mechanism predictions for the three surrogates (using different aromatic components) show no marked difference in ignition delay time. At high temperatures (greater than 1000 K), this mechanism and mixtures yield relatively good agreement with the current data. However, both these mechanisms fail to show any roll-off trend in activation energy below about 1050 K. Similarly, the Dagaut and Cathonnet mechanism and surrogate fail to predict ignition delay times for the 27

45 temperature range of current experiments at 20 atm pressure. It should be noted that both the Dagaut and Cathonnet and the Lindstedt and Maurice surrogate mixtures and their mechanisms contain species only up to C 10. The reasons for the failure of these two mechanisms could be due either to the chosen surrogate or the mechanism itself. It may be necessary to include higher molecular weight components, such as n-dodecane, which calls for significant modifications to be made in these mechanisms. Earlier, Holley et al. [109] suggested that the inclusion of small hydrocarbons in a surrogate will result in a fuel that may not mimic satisfactorily the flame behavior of real jet fuels. The Zhang et al. JP-8 mechanism predicts higher ignition delay times compared to the data for the temperature range studied at 20 atm, when modeled using the Violi #3 surrogate composition, and failed to capture the high-temperature (greater than 1000 K) trends in the activation energy. This mechanism was not designed for very low temperature ignition applications. Also, due to the absence of peroxy chemistry in this model, the mechanism does not show any roll-off behavior at low temperatures as observed in the data. The role of peroxy radicals in low-temperature combustion is significant and is described in detail by Miller et al. [111]. The Zhang et al. mechanism predictions using different surrogate fuel mixtures were analyzed. Predictions with neat n-dodecane were the fastest, since the n-alkane sub-mechanism used in this mechanism leads to shorter ignition delay times than the aromatic sub-mechanism. The Stanford B and Violi #3 surrogate mixture predictions did not differ at all when modeled using the Zhang et al. mechanism, suggesting that the aromatic sub-mechanism of the Zhang et al. mechanism is not complete. The Stanford A surrogate predictions showed the slowest ignition process, which suggested that adding more branched alkanes to the Violi #3 surrogate was not an option to improve the predictive abilities of the Zhang et al. mechanism. In general, both the Zhang et al. mechanism and the surrogate mixtures studied, fail to capture both the high temperature trends in ignition delay times and the low temperature behavior. 28

46 1250K 1000K 833K Ignition Delay Time [ s] Jet-A/air JP-8/air 100 All data scaled to 20 atm using P -1 Ranzi et al. Lindstedt et al. Zhang et al. Dagaut and Cathonnet /T [1/K] Figure 3.5 Jet fuel ignition delay times for phi=1 (data from Figure 3.3), and mechanisms prediction. The Violi #3 surrogate was used with the Ranzi and the Zhang et al. mechanisms. See text for details. The Ranzi mechanism gives the closest agreement with data under current conditions, when compared to other mechanisms (Figure 3.5). When modeled using the Violi #3 surrogate composition, predictions at 20 atm are slightly higher than the experiment at temperatures above 1000K. Below this temperature, the model shows strong NTC behavior due to the large fraction of n-dodecane in the surrogate fuel (details provided in [111]). In Figure 3.6, a comparison of the data with different surrogate mixtures using the Ranzi mechanism is shown. Peroxy chemistry plays a more important role in longer straight chain alkanes, and large n-alkanes ignite faster under milder conditions than small or branched chain alkanes. Thus, neat n-dodecane shows the strongest roll-off as expected (Figure 3.6). The prediction for the Violi #1 surrogate mixture shows the weakest low-temperature roll-off, more similar to the gradual roll-off trend seen in the data. However, the Violi #1 surrogate is relatively complex (6 components). Moreover, the ignition delay times predicted using the Violi #3 surrogate agree more closely with the data when compared to those using the Violi #1 surrogate at higher temperatures (> 1000 K). There is no significant difference between the predictions for the Stanford A and Stanford B surrogate mixtures. In general, the Ranzi mechanism captures the hightemperature trend, but fails to predict the low temperature behavior of ignition delay 29

47 times. Due to the simplicity of the Violi #3 surrogate mixture and the closeness to data when used with the Ranzi mechanism (compared to other surrogates at high T), this combination will be used in the later sections of this chapter in evaluating trends in experimental data. 1250K 833K All data scaled to 20 atm using P -1 Ignition Delay Time [ s] 1000 Jet-A/air JP-8/air 100 Violi et al. #3 Violi et al. #1 Stanford A Stanford B neat n-dodecane /T [1/K] Figure 3.6 Jet fuel ignition delay times for phi=1, and the Ranzi mechanism predictions for different surrogate fuel mixtures listed in Table 3.4. Further experimental work including key species time-history measurements is needed to provide more comprehensive kinetic targets for refining the mechanisms and surrogate mixtures. The current comparison study suggests that a better surrogate composition and/or improved model are particularly needed for simulating combustion behavior of jet fuels at low temperatures. With regard to improving the surrogate model, adding more aromatic compounds to the surrogate mixture might be one approach to slow down the chemistry in the low temperature region, so the mechanism predictions will not roll-off so early when compared to the data. Previously, many researchers have theoretically explained and experimentally confirmed the fact that addition of aromatics has an inhibiting effect on the chemistry of alkanes [21]. However, aromatics contribute significantly to pollutant formation, including soot, and thus have to be modeled with care. Moreover, it is important to note that pollutant or soot emissions often depend on trace fuel species and other additives present in the real fuel, and will not, in general, be reproduced by a simple chemical surrogate [6]. 30

48 3.3.5 Effect of Variation of Equivalence Ratio and Oxygen Concentration Ignition delay time correlations (τ ign as a function of T, P, X O2, and Ф) are widely used in codes for modeling engines, and by experimentalists to enable comparison of findings from different studies that might have been conducted with different reaction conditions [102]. However, due to the complexities associated with variations in jet fuel compositions, such an attempt will not be made in the present work. Instead, the results of jet fuel ignition delay time experiments with varying equivalence ratio and oxygen concentrations are presented separately in this section. Jet fuel experiments with equivalence ratios of Φ=1.0 and Φ=0.5 are compared in Figure 3.7. A clear trend of longer ignition delay times for the lean (Φ=0.5) as compared to the stoichiometric case (Φ=1.0) is noticed. Gauthier et al. [39] observed similar trends in gasoline/air mixtures for the same equivalence ratios. The reason for this trend with Φ is explained in detail by Curran et al. [113] in a comprehensive modeling study of oxidation using iso-octane. In effect, the higher the concentration of fuel in the mixture (for nearly constant O 2 concentration), the faster the ignition process become. Curran et al. [113] found that at temperatures below approximately 1150K, increasing Φ, i.e. increasing the fuel concentration, increases the alkyl-hydroperoxide radical pool production and results in shorter ignition delay times. After the initial fuel decomposition steps, the increase in the radical pool concentration is responsible for the rapid reactions associated with ignition. The Ranzi mechanism predictions using the Violi #3 surrogate mixture are also presented in Figure 3.7. The mechanism with this surrogate model consistently overpredicts the ignition delay times for both the lean and stoichiometric cases, but the trend with equivalence ratio is captured at higher temperatures (T >1000 K). Results of experiments in jet fuel/air mixtures with different oxygen concentrations (X O2 =0.10 and 0.207) are presented in Figure 3.7; the balance of gas mixture is N 2 in both cases. The low scatter data for 10 % O 2 clearly show longer ignition delay times, compared to the 20.7 % O 2 case. In both the low and high O 2 concentration cases, the fuel/air mixtures are stoichiometric. In effect, the lower the O 2 concentration the lower the fuel concentration, and following Curran et al. [113], the longer the ignition delay 31

49 times. The Ranzi et al. mechanism predictions using the Violi et al. #3 surrogate mixture (Figure 3.7) is able to capture the trend in the data for the effect of oxygen concentration on the ignition delay time, and the model predictions are in good quantitative agreement with the limited data for the 10% O 2 case. However, as mentioned in the previous section, the simulations do not recover the roll-off in ignition time observed at low temperatures for the 20.7% O 2 case K Data scaled to 20 atm using P K K Data scaled to 20 atm using P -1 10% O K Ignition Delay Time [ s] Jet-A/air, =1.0 JP-8/air, =1.0 Jet-A/air, =0.5 Ranzi et al. =1.0 Ranzi et al. = /T [1/K] Ignition Delay Time [ s] % O 2 Jet-A/air (20.7% O 2 ) JP-8/air (20.7% O 2 ) Jet-A/ 10% O 2 / N 2 Ranzi et al /T [1/K] Figure 3.7 Jet fuel ignition fuel ignition delay times: the effect of phi and X O Low-temperature (NTC Region) Ignition Delay Time Data The conventional shock tube driver gas is helium, since the strong incident shock waves desired for high test gas temperatures (T 5 ) are easily achievable with helium. But when helium is used as a driver gas, the available test times are typically limited to about 2 ms in the HPST. Available test time is defined as the time interval between the arrival of the reflected shock at the sidewall location (10 mm from endwall) and the arrival at that same location of a significant pressure disturbance, usually developed from an internal shock reflection from the contact surface or from a rarefaction wave propagating from the end of the driver section. Low-temperature ignition delay time data requires longer test times than are normally available with conventional shock tube operation. To increase our shock tube test times, the driver and driven gas mixtures were tailored to allow study of the combustion chemistry at low-to-intermediate 32

50 temperatures (see, for example, Amadio et al. [114]). Using pure N 2 as the driven gas, a series of experiments were performed with different He/N 2 driver gas mixtures (20-30 % N 2 in He) in the low temperature region from K in the HPST (for P 5 of approximately 20 atm). Test times in excess of 4 ms were readily achieved. Sample pressure traces for a non-reactive (N 2 ) case and ignition traces for a reactive (Jet-A/air) case are shown in Figure 3.8, along with the OH* record. 80 jet-a/air, = K, 19.4 atm K 909K 769K 60 Pressure [atm] Ignition time = 2570 s pressure OH* Pure N K, 19.0 atm Time [ s] Ignition Delay Time [ s] P~20atm jet-a/air, =1.0 Ranzi (Violi 3 surrogate) /T [1/K] Figure 3.8 Left: low-temperature pressure and emission traces (Driver gas: He 70%, N 2 30 %). Right: ignition delay times including NTC region data, (Jet-A/air, phi=1). The ignition delay time variation with initial reflected shock temperature, normalized to 20 atm (assuming τ ign ~ 1/P), is presented in Figure 3.8. At these low temperatures, ignition delay times are usually long (> 2 ms) and commonly show NTC type behavior, in which the ignition delay time remain nearly constant or become shorter as the temperature decreases. The Ranzi mechanism prediction using the Violi et al. #3 surrogate mixture, rolls off much earlier, but later catches up, retaining the primary feature of the data. At very low temperatures (below 700K), ignition delay times again increase with decreasing temperature, as seen both in the data and the model results. The pressure-time history data below 800 K in the NTC region show mild pre-ignition heat release in Jet-A/air mixtures at pressures close to 20 atm. This is evident in the pressure and emission traces for the case shown in Figure 3.8 (seen as slow, exponential growth to strong ignition). Since the test gas behind reflected shock waves behave like a constant volume reactor for weak to moderate energy release, pre-ignition pressure rises are related to rapid changes in chemistry with relatively small concomitant energy release 33

51 (Davidson and Hanson [102]). In the current experiments, all the pressure-time history data look similar to that in Figure 3.1 for temperatures higher than 800 K. Modeled pressure traces, by the Ranzi mechanism using the Violi et al. #3 surrogate mixture (not shown here), do not show this slow, exponential growth in pressure. Nevertheless, for temperatures below 900 K, the Ranzi mechanism predictions show a two-stage ignition behavior (pressure smoothly rises to an intermediate plateau before ignition). In the case of iso-octane/air mixtures, Davidson et al. [115] suggested the importance of using accurate values of the heats of formation and rate coefficients of various peroxy species, which dominate the pre-ignition heat release stage of combustion. For long test times, at a fixed location from the end-wall of the shock tube, and in the absence of reaction, the reflected shock pressure gradually increases, approximately linearly with time. This is caused by non-ideal effects such as incident shock attenuation, boundary layer growth, and interaction of the reflected shock wave with the side wall boundary layer. Petersen and Hanson [116] confirmed that an isentropic relation between pressure and temperature can be used to estimate the temperature increase from pressuretime histories in the same facility. Comparison between pressure increases in nonreactive (pure N 2 ) and reactive (Jet-A/air, Φ=1.0) mixtures, shows a higher pressure increase with time in the latter case due to the pre-ignition energy release caused by chemical reactions. In the non-reactive case, for the temperature range from K at a pressure close to 20 atm, a simple relation could be established for the reflected shock pressure increase and the incident shock attenuation. At the current side-wall location (10 mm from end-wall), a linear relation between change in reflected pressure with time (dp/dt) and the incident shock attenuation rate (%/m). This relation could be used to estimate the total temperature increase due to non-ideal effects from the arrival of reflected shock wave till the onset of ignition, even in the case of reactive mixtures. In the current experiments for Jet-A/air ignition delay times below about 4 ms, the maximum increase in temperature was estimated to be only 21 K (using the isentropic assumption). However, as seen in Figure 3.8, it is to be noted that ignition delay times are relatively insensitive to small variations in the temperature in the NTC region as the ignition delay times in this region are nearly constant. Hence the constant U, V modeling is justified in this special case to model data. 34

52 It is clear that additional studies of surrogate blends, as well as studies of the effects of single reference components on the mixture ignition, are needed to support the jet fuel surrogate development. In the subsequent chapters of this thesis, major single reference components present in jet fuels will be studied to understand their ignition chemical kinetics and the measurement of species concentration time-histories will be conducted, particularly, for important radicals such as OH that can be used to validate assumptions made about the internal structure and sub-mechanisms of these large reaction mechanisms. 35

53 Table 3.1 Summary of Jet-A/air ignition data at phi=1.0. Jet-A Composite Blend #04POSF4658. (Dr. = driver gas mixture; V shock = incident shock velocity at the endwall). Jet-A/air, phi=1.0, Jet-A=1.276 %, O 2 =20.74 %, N 2 =77.98 %, Dr.= He P 1, (psi) T 1, (ºC) V shock,(mm/us) T 5,(K) P 5, (atm) ign, (us) Jet-A/air, phi =1.0, Jet-A=0.615 %, O 2 =10 %, N 2 =89.38 %, Dr.= He Jet-A/air, phi =1.0, Jet-A=1.276 %, O 2 =20.74 %, N 2 =77.98 %, Dr.= He/N

54 Table 3.2 Summary of jet fuel ignition data indicating variation with fuel type (jet-a and JP-8), equivalence ratio, and fuel sample. (Dr. denotes driver gas mixture; V shock is the incident shock velocity at the endwall). JP-8/air, phi =1.0, JP-8=1.276 %, O 2 =20.74 %, N 2 =77.98 %, Dr.= He P 1, (psi) T 1, (ºC) V shock,(mm/us) T 5,(K) P 5, (atm) ign, (us) Jet-A/air, phi =0.5, Jet-A=0.642 %, O 2 =20.87 %, N 2 =78.48 %, Dr.= He (Jet-A Composite Blend #04POSF4658) Jet-A/air, phi =1.0, Jet-A=1.276 %, O 2 =20.74 %, N 2 =77.98 %, Dr.= He (Jet-A fuel same as Dean et al. [9])

55 Table 3.3 Jet fuel surrogate mechanisms and their associated surrogate mixtures. Kinetic Mechanism, Surrogate Mixture Surrogate Mixture Composition Lindstedt and Maurice [112] Dagaut and Cathonnet [8] n-decane 89% benzene, (or) toluene, (or)ethyl benzene (11%) n-decane 74% n-propylbenzene 15% n-propylcyclohexane 11% Table 3.4 Jet fuel surrogate mixtures used with the Ranzi [48] and the Zhang et al. [19] mechanisms. Kinetic Mechanism Ranzi [48], Zhang et al. [19] Surrogate Mixture Violi #1 (Violi et al. [5]) Violi #3 (Violi et al. [5]) Stanford A Stanford B Surrogate Mixture Composition MCH 20% m-xylene 15% tetralin 5% iso-octane 10% n-dodecane 30% tetradecane 20% MCH 10% toluene 10% benzene 1% iso-octane 5.5% n-dodecane 73.5% MCH 10% toluene 10% benzene 1% iso-octane 25% n-dodecane 54% MCH 10% toluene 29.5% benzene 1% iso-octane 5.5% n-dodecane 54% 38

56 Chapter 4: High-pressure Single- Component Surrogate Ignition Times This chapter describes ignition delay time measurements in single-component jet fuel surrogates using the heated HPST. Normal and cyclo alkanes are the two most important chemical classes found in jet fuels as mentioned in the previous chapter. Accordingly, high-pressure ignition experiments are conducted in two commonly used representative components for normal and cyclo alkanes in jet fuel surrogates (i.e., n-dodecane and MCH, respectively). n-dodecane/air ignition was studied for the following shock conditions: temperatures of K, pressures of atm, equivalence ratios of 0.5 and 1. Experimental conditions in MCH/air (phi=1.0) were in the range T= K and P= atm. Detailed comparisons of experimental data with predictions of available n-dodecane and MCH mechanisms are also presented. 4.1 Experimental Method All experiments were carried out in the reflected shock region of the HPST. The method and diagnostics were similar to that described in previous chapters. Helium was used as the driver gas in most cases except to access the low-temperature regime ( K), where tailored driver mixtures of 20-30% N 2 in He were used. In the case of n- dodecane, mixtures of research grade ( 99%) n-dodecane (Sigma-Aldrich) and highpurity grade(> %) gases (Praxair N 2, O 2, and Ar) were prepared. The shock tube was heated to 105 C for experiments to measure ignition times in synthetic air (79% N 2, 21% O 2 ), and the fuel/oxidizer mixing facility was kept at 135 C in both cases. This heating prevented condensation of the fuel during the mixing and shock tube filling procedure. The shock tube can be heated to lower temperatures as the partial pressure of n-dodecane is lower than in the mixing tank and care was taken to verify that this method has no effect on kinetics findings. Mixtures of research grade ( 99.5% pure) MCH (from Sigma-Aldrich) and high-purity synthetic air (Praxair 79% N 2 and 21% O 2, >99.999%) were prepared. In the MCH experiments, the mixing tank and connecting gas lines to the shock tube were heated to 110 C (where the fuel is at a higher partial pressure) and the 39

57 shock tube was used both in an unheated and heated (105 C) configuration (where the pre-shock partial pressure of the fuel is lower) to check the influence of shock tube heating on the measured ignition delay times. The onset of ignition from the pressure history and OH* emission were defined by locating the time of steepest rise and linearly extrapolating back in time to the preignition baseline. The two methods give very similar results (± 1%) and ignition delay times are readily identifiable with both diagnostics (see Figure 4.1 for a sample ignition data from HPST for n-dodecane and MCH). The overall uncertainty in τ ign was ±10%, dominated by uncertainties in the determination of T 5 (±1%). Pressure, [atm] n-dodecane/air =0.5 T 5 =996K P 5 =20.3 atm Pressure You et al. Ranzi et al. Pressure, [atm] MCH/air, =1.0 T=918 K, P=46.4 atm ign =1185 s Pressure 0 ign =1245 s OH*emission 0 OH* emission Time, [ s] Time, [ s] Figure 4.1 Example ignition data from HPST. Left: n-dodecane/air. Right: MCH/air. In the case of n-dodecane, comparisons with predictions of four recent detailed kinetic mechanisms, one n-dodecane mechanism and three JP-8 mechanisms where n-dodecane is an important surrogate component are provided: Ranzi [48] JP-8 semi-detailed mechanism (280 species, 7800 rxns.); Zhang et al. [19] JP-8 detailed mechanism (208 species, 1087 rxns.); Montgomery et al. [117] reduced JP-8 mechanism (94 species, 675 rxns.); and the You et al. [118] (which is now part of the JetSurF 1.0 [46] surrogate mechanism) n-dodecane detailed mechanism (177 species, 1318 rxns.). Similarly, in the case of MCH, comparisons are made with the predicted ign using three recent mechanisms: 1) the Pitz et al. [22] MCH mechanism, 1001 species, 4436 rxns., 2) the Orme et al. [36] MCH mechanism, 190 species, 904 rxns., 3) the Ranzi [48] JP-8 surrogate mechanism, 280 species, 7800 rxns. The kinetic implications of these 40

58 comparisons, particularly in the low-temperature negative-temperature-coefficient (NTC) region where the role of non-ideal facility effects can be significant, are discussed. 4.2 n-dodecane/air Ignition: Results and Discussion High-pressure n-dodecane Ignition Data The ignition delay time measurements and modeling for n-dodecane/air for phi=1 and 0.5 are shown in Figure 4.2. All data are summarized in Table 4.1 and Table 4.2. As the current study is over a limited pressure range (mainly atm), data are normalized to 20 atm using a τ ign ~ 1/P scaling in all figures (unless mentioned otherwise). As discussed in the previous chapter, this is a good first approximation for jet fuels and surrogate components for similar fuel concentrations, temperature and pressure. However, the actual pressure dependence of τ ign is expected to vary with temperature, particularly at the lower temperatures, and further studies are needed to more completely characterize this dependence. All data points are plotted in Figure 4.2 at the value of T pertaining immediately after reflected shock heating. The data are characterized by small scatter and show a non-monotonic variation with temperature. At high temperatures (above ~ 1000 K) ignition delay times were observed to decrease as temperature is increased. However, at low temperatures, τ ign shows NTCtype behavior, in which τ ign remains nearly constant or decreases as the temperature decreases. A distinct variation of ignition delay time with equivalence ratio (phi) is also seen. Ignition delay times for the lean ( =0.5) case are consistently longer than for the stoichiometric measurements ( =1.0). 41

59 1176 K 1000 K 833 K 714 K n-dodecane/air 20 atm Ignition Delay Time, s] =0.5 =1.0 You et al. Ranzi et al. Montgomery et al. Zhang et al /T, [1/K] Ignition Delay Time, s] 1176 K K 833 K 714 K n-dodecane/air 20 atm, =1.0 Current Study You et al. (2007) Zhang et al. (2007) Ranzi et al. (2006) Montgomery et al. (2007) /T, [1/K] Figure 4.2 n-dodecane/air τ ign data (20atm) and predictions for =0.5 (left) and1.0 (right) Kinetic Modeling Prediction The current data set provides critical validation targets for n-dodecane and jet fuel surrogate mechanisms. Ignition time predictions of the detailed You et al. [118] mechanism along with the two semi-detailed (partially lumped) JP-8 mechanisms of Ranzi [48] and Zhang et al. [19] (Zhang et al. mechanism was not optimized for low temperature modeling), and an optimized reduced mechanism of Montgomery et al. [117], are presented in Figure 4.2 using the constant U,V calculation method (CHEMKIN [101]). Both the Ranzi and the Montgomery et al. mechanisms predict similar results, and give relatively close agreement in magnitude with data (at both phi s), and are able to qualitatively capture the negative temperature coefficient-type (NTC) roll-off seen in data. As discussed in the previous chapter, the Ranzi mechanism applied to the Violi et al. [5] JP-8 surrogate mixture (73.5% by volume of this surrogate is n-dodecane) predicts a rolloff trend similar to that seen in our measured ignition delay times near 20 atm for Jet- A/air mixtures. However, in that case, the Ranzi mechanism predicts stronger roll-off than observed in Jet-A/air data, consistent with the current situation for n-dodecane (Figure 4.2). The You et al. detailed n-dodecane mechanism is less successful overall but provides reasonably good agreement with the data at T above 840K for both equivalence ratios. Note that all four mechanisms show little difference in simulated ignition delay times between the two phi s at the highest temperatures of the current study (near 1200 K), for the 20 atm pressure condition. 42

60 Sensitivity analysis at 20 atm for stoichiometric n-dodecane/air using the You et al. mechanism confirms the importance of RO 2 chemistry below about 1000K during preignition and the importance of n-dodecane peroxy and hydro-peroxy radical isomerization chemistry during the initial fuel breakdown steps during the induction time (specifically at 833K). At high temperatures (1200K), H 2 O 2 chemistry plays a dominant role in n-dodecane/air ignition Effect of Pressure-time Histories To simulate the potential effect of facility-related pressure rise on measured and modeled τ ign, a CHEMSHOCK [103] calculation with the You et al. mechanism using an extreme case of a facility-dependent pressure rise of 10%/ms (see Figure 4.3). For the current HPST experiments, the observed dp 5 /dt was typically between 1 to 10% /ms (which may also include the effect of pre-ignition energy release due to chemistry). From Figure 4.3, it is evident that there is no significant difference between the constant U, V model calculation and the CHEMSHOCK model calculation, which includes this facilitydependent dp 5 /dt, for temperatures above 1000 K. At temperatures between about 1000 and 850 K, the fluid-mechanically more accurate CHEMSHOCK model predicts ignition times that are up to 30% longer than the constant U, V model and results in a model prediction that is in better agreement with current data. The disagreement between the CHEMSHOCK (and constant U,V) simulations and the measured ignition delay times below 800K is currently attributed to uncertainties and limitations of the kinetic mechanism. Specifically, below 800 K, improved knowledge of the NTC chemistry (such as the peroxy isomerizations) is needed. Evidence of pre-ignition pressure rise is seen at low temperatures in the data set. In Figure 4.3, the measured pressure profiles below 900 K show a perceptible pre-ignition induction time, then a step change to an intermediate plateau, and then finally, a second large rise at the point of actual ignition. This pre-ignition induction time appears to be longer at both high (869 K) and low (727 K) temperatures and has a minimum in the range K. Additionally, the relative magnitude of this pre-ignition pressure step increases (almost linearly) as temperature decreases. This behavior differs from the measured pressure profiles of Jet-A (previous chapter), which show only mild pre- 43

61 ignition heat release at temperatures below 800 K. Modeled pressure traces using the You et al. mechanism (not shown here), also show this pre-ignition behavior at temperatures below 975K for Ф=1 at 20 atm for n-dodecane/air. Ignition Delay Time, s] 1176 K 1000 K 833 K 714 K n-dodecane/air =1.0 Current Study You et al. CHEMSHOCK You et al. Const U,V /T, [1/K] Pressure, [atm] K 727K 869K 822K 957K n-dodecane/air, = Time, [ s] Figure 4.3 Left: n-dodecane /air τ ign results for phi=1.0. CHEMSHOCK modeling was conducted using the You et al. mechanism with an extreme-case linear pressure rise (dp 5 /dt) of 10% per milli-second. Right: Measured n-dodecane/air pressure-time histories near 20 atm, phi= Comparison of n-dodecane Ignition Times with Jet-A and Other n-alkanes The current τ ign measurements for n-dodecane/air are compared with the τ ign data for Jet-A/air acquired in the previous chapter and are shown in Figure 4.4. At temperatures above about 950 K and for both phi s, τ ign of these fuels do not show significant differences. Below this temperature, the n-dodecane ignition times are shorter (compared to Jet-A), and this difference is largest for the case of phi=1. Thus a multi-component jet fuel surrogate (instead of neat n-dodecane) would likely be needed to accurately reproduce the τ ign data in the NTC region (< 950K), as suggested in the previous chapter. However, these results also imply that using n-dodecane as a single component surrogate for jet fuels may be adequate to represent high-temperature combustion conditions. This similarity between n-dodecane and jet fuels has been observed earlier in other types of experimental studies: Eigenbrod et al. [14] observed the similarity between n-dodecane 44

62 and kerosene droplet induction times, and Holley et al. [16] found that extinction strain rates of n-dodecane and jet fuel non-premixed flames are very close. Comparison of 3 different n-alkane ignition delay times are also presented in Figure 4.4 (all data scaled to 20 atm using τ ign ~1/P dependence) for phi=1.0 in air. n-heptane ignition times in the NTC region near 20 atm were measured in the range K in this work in order facilitate comparisons between n-alkanes for a range of temperatures. Note that Figure 4.4 includes measurements in n-decane by Pfahl et al. [119] (which was at 12.8 atm and was not from the HPST) and previous high temperature n-heptane data from Gauthier et al. [39]. Omitting the fact that the n-decane measurements near 850 K are close to those of n-heptane, in the entire temperature regime, n-heptane ignition times are consistently longer than those of n-decane (which in turn is longer than those of n- dodecane). A general trend is clearly seen: there is a slight trend where the ignition delay time increases as the length of the n-alkane chain decreases at post-shock temperatures in the entire temperature regime from ( K). Within the experimental uncertainties, the above observation is not consistent with the modeling conclusions made recently by Westbrook et al. [47] (also shown in Figure 4.4). Westbrook et al. [47] concluded that in a shock tube at stoichiometric conditions, any of these n-alkanes from n-heptane to n-hexadecane can be used to substitute for any other, with equivalent results. They also cited the success of n-heptane as a model diesel surrogate (despite its smaller size than conventional diesel components); n-heptane ignition rate is close enough to those of real diesel fuels, whose ignition rate and cetane number are established by its large n-alkane components, that diesel ignition is quite well reproduced by n-heptane. However, current work (Figure 4.4) indicates the following: n- dodecane is not a good surrogate for jet fuels in the NTC region, n-heptane similarity with diesel might be fortuitous, and calls for caution while making kinetic conclusions using simple surrogate to represent practical fuels. 45

63 1250 K 1000 K 833 K 714 K Ignition Delay Time, s] Jet-A/air, =0.5 Jet-A/air, =1.0 n-dodecane/air, =0.5 n-dodecane/air, = /T, [1/K] Figure 4.4 Comparison of high-pressure n-dodecane ignition delay times at 20 atm. Left: with Jet-A/air at two equivalence ratios. Right: with n-heptane/air (current work) at =1.0. Dashed lines are fit through data. Solid lines are Westbrook et al. [47] (LLNL) modeling prediction. 4.3 MCH/Air Ignition: Results and Discussion High-pressure MCH Ignition Data All high-pressure τ ign data for MCH/air (Mixture: MCH=1.96%, O 2 =20.60%, N 2 =77.44%, =1.0) in the range T= K and P= atm are summarized in Table 4.3. The high-pressure τ ign data for two pressures near 20atm and 45atm, both for an equivalence ratio of 1, are plotted in Figure 4.5. In all figures, data labeled as heated refers to experiments when the shock tube was kept at 105 ºC, otherwise the shock was kept at room temperature (T 1 = 22.5 ºC). There does not appear to be a significant difference in τ ign measurements when the shock tube was heated to T 1 = 105 ºC or when shock tube was kept at room temperature (T 1 = 22.5 ºC). This also supports our observation from the previous chapter, where ignition delay time variation with T 1 in the case of jet fuel was found to be negligible. Current data are characterized by small scatter and we have used the experimentally found (τ ign ~ P ) relation to scale τ ign to nominal pressures in Figure 4.5. It should be noted that the pressure dependence of the ignition delay time is expected to vary with temperature and researchers have used different pressure scaling in different temperature regimes for similar hydrocarbons [120]. Accordingly, the current data follows τ ign ~ P dependence above 912K (without 46

64 including the NTC region data points). Due to the changing activation energy in the data, a standard Arrhenius regression expression of current data is not possible, however, in the high-temperature region above 912K, an overall activation energy of 24.9 kcal/mol was obtained using the Arrhenius expression. In Figure 4.5, at low temperatures (less than about 880 K), ignition delay times show negative-temperature-coefficient-type (NTC) behavior. 1667K 1000K 769K 1250K 1000K 833K 714K Ignition Delay Time, [ s] MCH/air 45atm 45atm (heated HPST) 20atm (heated HPST) 1.5atm data MCH/air, = Modeling, 20atm Orme et al. Ranzi et al. Pitz et al /T, [1/K] Ignition Delay Time, [ s] MCH/air, 45atm, =1.0 Orme et al. Ranzi et al. Pitz et al /T, [1/K] Figure 4.5 Left: High-pressure MCH/air ignition delay time results ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%). High-pressure data scaled to 20, or 45 atm using τ ign ~ P Low-pressure data scaled to 1.5atm using Vasu et al. [121] correlation ( =1.0, MCH=1.962 %). Modeling shows τ ign predictions at 20 atm. Right: High-pressure MCH/air ignition delay time results near 45 atm ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%) and pressure scaling. Data (solid symbols) scaled to 45 atm using τ ign ~ P Grey solid line is a fit through data. Constant U,V modeling results at 20 and 45 atm are shown using Orme et al., Ranzi et al., and Pitz et al. mechanisms. Also shown in Figure 4.5 are the low-pressure data from the Vasu et al. [121] study extrapolated to the conditions of 1.96% MCH, phi = 1.0, 1.5 atm. Current data have significantly extended the range of high-temperature MCH ignition delay times. Also, the near-unity pressure dependence (P -1 ) of the lower-pressure data (P from Vasu et al. [121] empirical correlation) and the current higher-pressure data (P -0.87) extends up to 1560 K. Additionally, current results complement earlier ignition time studies of MCH by Pitz et al. [22] in a RCM (see Figure 4.6) and the combined data sets provide kinetic 47

65 targets in the range from K. The RCM τ ign are less reproducible (have more scatter than the current high-pressure data) when time scales reach near 50ms; see, for example, the observations of Dooley et al. [122]. By examining all the data from 45 to 10 atm, it is evident that in the intermediate temperature range near 850 K, the pressure appears to have the most pronounced influence on the measured ignition delay times. However based on the RCM measurements, at low temperatures near 700 K, ignition times are almost independent of pressure. 1250K 1000K MCH/air, = K 714K 625K K 1000K = K 625K Ignition Delay Time [ s] /T [1/K] 3 current data 20 atm (heated) 45 atm 45 atm (heated) Pitz et al. RCM 10 atm 15 atm 20 atm Ignition Delay Time [ s] atm MCH/air current data 20 atm (heated) 45 atm 45 atm (heated) 100 Pitz et al. RCM 10 atm 15 atm 20 atm /T [1/K] Figure 4.6 Current high-pressure MCH/air ( =1.0, MCH=1.96%, O 2 =20.60%, N 2 =77.44%) τ ign results. Left: Current data scaled to 20, or 45 atm using τ ign ~ P (-0.87). RCM data (unheated) from Pitz et al. [22]. Lines are constant U,V modeling predictions using the Pitz et al. [22] mechanism: 1) 10 atm; 2) 20 atm; and 3) 45 atm. Right: HPST data scaled to 20, or 45 atm using τ ign ~ P (-0.87) ; RCM data (unheated) from Pitz et al. [22]; Solid lines are fit through data at respective pressures. Low-pressure data scaled to 1.5atm using Vasu et al. [121] correlation ( =1.0, MCH=1.962 %). It should be noted that the Pitz et al. [22] RCM measurements used 3 different diluents (100% N 2, 50%N 2 :50% Ar, and 100% Ar) to access their temperature range. According to a recent study by Würmel et al. [123], the choice of diluent gas affects RCM τ ign measurements. Würmel et al. reported that the effect of adding argon to nitrogen (changing the diluent gases is a standard procedure used to achieve various compressed temperatures in most RCM facilities) is to decelerate ignition in a RCM at the same compressed temperature and fuel and oxygen concentration compared to pure N 2 as 48

66 diluent. This increased ignition time is due to the extreme cooling of Ar in the postcompression period and could be attributed to the effect of the heat capacity of the bulk carrier gas (see Würmel et al. [123] and Davidson and Hanson [102]). Such an effect could explain the slightly higher activation energy at high-temperatures (above 833 K) shown by the RCM data at 10 atm than the current data at 20 atm and 45 atm (see Figure 4.6). Another observation from Figure 4.6 is that as the temperature is decreased, occurrence of NTC behavior (roll-over) is only slightly delayed, but is stronger, at lower pressures near 10atm than at 45atm Comparison of Various Cyclo-alkanes Daley et al. [124] recently measured τ ign in cyclo-pentane/air and cyclo-hexane/air mixtures at high pressures and high temperatures in a shock tube (their conditions are similar to current experiments in MCH). A comparison of the high-pressure shock tube τ ign results for three cyclo-alkanes/air are shown in Figure 4.7 (for 45atm and =1.0). Current MCH pressure scaling of τ ign ~ P and Daley et al. [124] pressure scaling of τ ign ~ P -1.1 for cyclo-hexane and τ ign ~ P -0.9 for cyclo-pentane were used in Figure 4.7. It is interesting to note that the activation energies for these three cyclo-alkanes are nearly the same, i.e, 23.9 kcal/mol for cyclo-pentane; 24.9 kcal/mol for MCH; and 27.6 kcal/mol for cyclo-hexane. τ ign in MCH falls between the data for the other two cyclo-alkanes, and cyclo-hexane τ ign values are approximately half of those for cyclo-pentane. This suggests that reactivity of these fuels is in the order cyclo-hexane>mch>cyclo-pentane, which could be attributed to the relative stability of primary cyclo-alkyl radicals formed from these fuels and their propensity to yield H-atoms (see Sirjean et al. [125] for details on the high-temperature reactivity of cyclo-alkanes). However, it should be noted that the above observation is true only in the high-temperature region (above 900K), where the influence of NTC chemistry is minimal. It is clear from Figure 4.7 that MCH shows NTC-type behavior at higher temperatures (τ ign data roll-off at a higher temperature) than compared to cyclo-hexane at 45 atm. This indicates that propensity for RO 2 radical isomerization (and thereby NTC behavior) is higher in MCH and that MCH would be more reactive in the NTC region than cyclo-hexane. Similar conclusions in the NTC 49

67 region can be reached by comparing the RCM ignition delays of MCH (Pitz et al. [22]) and cyclo-hexane (Lemaire et al. [126]), both at pressures near 10atm. Additionally, at high temperatures and low pressures the activation energy obtained in argon-dilute mixtures (see Vasu et al. [121]) is almost two times that observed at high pressures and moderate temperatures (such as results in this section) with air as the oxidizer. Daley et al. [124] observed a similar difference in activation energy, consistent with most hydrocarbon fuels in the literature (such as toluene, n-heptane, iso-octane, and cyclo-pentane), and attributed this difference due to influence from the NTC chemistry in high-pressure, moderate-temperature studies atm, =1.0 fuel/air Ignition Delay Time, [ s] MCH (current) cyclohexane (Daley et al.) cyclopentane (Daley et al.) /T, [1/K] Figure 4.7 High-pressure cyclo-alkanes/air ( =1.0) τ ign results. Mixtures: MCH=1.96%, O 2 =20.60%, N 2 =77.44%; cyclo-hexane=2.28%, O 2 =20.53%, N 2 =77.19%; cyclopentane=2.72%, O 2 =20.44%, N 2 =76.84%. Current MCH data scaled to 45 atm using τ ign ~ P (-0.87). Cyclo-hexane (scaled using τ ign ~ P -1.1 and cyclo-pentane (scaled using τ ign ~ P ) data from Daley et al. [124]. Solid lines are fit through data Kinetic Modeling Model predictions using the mechanisms of Ranzi [48], Pitz et al. [22] and Orme et al. [36] are presented in Figure 4.5 for 20 atm and 45 atm, respectively. The Ranzi mechanism gives the closest agreement and the Orme et al. mechanism predictions are the farthest from data. In general, all mechanisms predict longer ignition delay times than experimental results, and none of the mechanisms except the Pitz et al. mechanism 50

68 exhibits the NTC roll-off behavior. However, at the peak near 870 K (before the NTC roll-off starts in both the 45 atm data and the Pitz et al. model), ignition delay times predicted by the Pitz et al. mechanism are approximately 5 times larger than data. At temperatures higher than 912 K, the global activation energies for ignition and pressure dependence according to various mechanisms are as follows: 34.0 kcal/mol and τ ign ~ P (Orme et al.), 29.5 kcal/mol and τ ign ~ P (Pitz et al.), and 32.9 kcal/mol and τ ign ~ P (Ranzi), whereas the experimental values are 24.9 kcal/mol and τ ign ~ P Compared to the other two mechanisms, the Pitz et al. mechanism provides slightly better agreement with the experimental pressure dependence and activation energy. However, the Pitz et al. mechanism does not predict their RCM measurements very well (see Figure 4.6) using the constant U,V approach, which was used also by Pitz et al. to develop their mechanism by validating against their RCM data. It should be kept in mind that the constant U,V approach may be incorrect for RCM environments and various other approaches (according to Würmel et al. [123], there is currently no off-the-shelf simulation tool available that allows the realistic description of combustion in a RCM) have been used in the literature to model RCM experiments mainly due to the heat loss and compression stroke effects in RCMs (see [26,123] for details on this topic). Since peroxy chemistry is the dominant channel in the NTC region, current observations (Figure 4.5) suggest that the MCH sub-mechanism included in the current detailed Ranzi [48] JP-8 mechanism needs to be improved (in order to accurately model the current MCH and Jet-A data in the NTC region) by adding or modifying reactions for the peroxy reaction channels of the MCH oxidation. Of considerable importance is the finding by Pitz et al. [22] that the RCM ign data in the NTC regime for MCH, where MCH oxidation proceeds via the peroxy channels, was found to be greatly influenced by the calculated RO 2 (here methylcyclohexylperoxy radical) isomerization rates. Specifically, the Pitz et al. computations using n- and iso-alkane-based estimates of methylcyclohexylperoxy radical isomerization rate constants predicted ignition delay times too long compared to their experiments. Ignition delay time sensitivity analyses were performed using both the Orme et al. [36] and the Pitz et al. [22] mechanisms. The key reactions that influence the ignition times both at 910K and at 1110K (for P=20atm) are nearly the same. Specifically, the formation 51

69 (via HO 2 +HO 2 =H 2 O 2 +O 2 ) and decomposition (via H 2 O 2 +M=OH+OH+M) are the dominant reactions influencing τ ign. Efforts to adjust the mechanisms by modifying these two reactions k 1 and k 2 within their uncertainty limits were not conducted; however, the Tsang and Hampson [127] review estimated an uncertainty factor of 5 for the first reaction, and very few experimental data exist for the latter one. Additionally, H 2 O 2 formation via H-abstraction reactions from MCH by HO 2 radical, i.e., MCH+HO 2 =H 2 O 2 + methylcyclohexyl isomers (3 isomers formed at primary, secondary, and tertiary sites), have considerable impact on the predictions of mechanisms, and these rate constants only have been estimated and not measured. Hence, the above mentioned reactions should be measured directly in order to enable accurate modeling of MCH ignition times at engine-relevant conditions (such as those presented in the current chapter). The low-pressure limit of the H 2 O 2 +M=OH+OH+M reaction has been recently measured in our lab [128], however, adopting this value into the Pitz et al. [22] mechanism did not improve τ ign predictions at 45 atm Effect of Pressure-time Histories The measured pressure profiles at some conditions, including Figure 4.8, show a weak, approximately linear pressure ramp starting just before ignition, which is defined as the ratio DP/Dt in Figure 4.8 (note that this definition of pressure rise in different from dp/dt, which was defined in the previous chapter). This type of behavior has been observed by others in various other fuels in high-pressure shock tube ignition studies [38,40,115,119, ] and in RCM ignition studies [132,133] and is currently a topic of intense research. Note that there is no pre-ignition rise in measured OH* emission signal (see Figure 4.1) for this case. Various hypotheses [115,129,132, ] for this pressure increase (DP/Dt) have been advanced in literature including chemical heat release before ignition (either at the measurement location or elsewhere in the reflected shock region), different modes or regimes of ignition, multi-dimensionality of ignition, etc. The pre-ignition phenomena (or DP/Dt) may or may not be a result of homogeneous chemistry alone (here homogeneous refers to uniformity at a given axial location, not necessarily the entire reflected shock zone), and could be a result of various non-uniform phenomena (such as axial or even 3-D inhomogeneities which might arise in temperature, 52

70 radical concentration, or particles in the test gas) acting in addition to chemical heat release [131,135]. In the shock tube, the main cause of temperature inhomogeneities in the reflected shock region is likely to be the interaction of the reflected shock wave with the boundary layer arising from the incident shock [129,142]. Also, in the current experiments at high-pressures and all temperatures, ignition events could be classified as strong ignition [134,138], which is characterized by large pressure peaks immediately after the main ignition event and by the following large-amplitude pressure oscillations seen in Figure 4.8. These observations are consistent with strong ignition observations in n-heptane/air [38] and in iso-octane/air [129] in other shock tube ignition studies and is not unique to current facility. A detailed discussion on this topic is provided in Appendix. However, simple constant U,V calculations do not capture the experimentally measured pressure oscillations caused by the blast wave after ignition (Figure 4.8). For the current MCH experiments, some qualitative observations can be made about the nature and magnitude of DP/Dt. Figure 4.9 provides pressure profiles from HPST experiments for different temperatures (T 5 ) and 2 pressure regions (20 and 45 atm), illustrating that the DP/Dt values do not show any strong dependency on pressure (i.e., no variation with P 5 when T 5 is constant). In addition, it was observed that DP/Dt values are slightly lower for heated shock tube experiments for the same P 5 and T 5 compared to room temperature experiments (not shown here). As T 5 increases from 795 K, DP/Dt values remain approximately constant until about T 5 =950 K, and increase after that until T 5 =1100 K. Consequently, variation of DP/Dt with τ ign follows the opposite trend, i.e., with increasing τ ign, DP/Dt decreases gradually until τ ign = 600 s and remains nearly independent of τ ign thereafter till τ ign = 2ms (the highest τ ign for current experiments). However, at 795K (the lowest temperature near 45 atm, which is labeled as 5 in Figure 4.9) the pressure starts to behave more like a two-stage ignition than showing a linear DP/Dt as defined in Figure 4.8. It should be also noted that above 950 K, the duration (Dt) of pressure ramp decreases with increasing temperature in MCH/air mixtures (current experiments), and this trend is consistent with observations made by Pfahl et al. [119] in their high-pressure shock tube experiments performed in stoichiometric α- methylnaphthalene/air mixtures at 12.8 atm. 53

71 750 MCH/air, =1.0, T 5 =918 K, P 5 =46.4 atm 750 Pressure [atm] HPST data Pressure [atm] DP Time [ s] Pitz et al. Ranzi et al. Dt Presure, [atm] Pitz et al. mechanism MCH/air, =1.0 20atm 775 K 725 K 875 K K Time [ s] Time, [ms] Figure 4.8 Left: Measured and predicted (Ranzi et al. [48], Pitz et al. [22]) P(t) histories. Right: Computed pressure-time histories for MCH/air ignition at 20atm, =1.0, Pitz et al. [22] mechanism K Pressure, [atm] [A] 1 3 MCH/air, = Time, [ s] 4 1) 1029K 2) 1019K 3) 962K 4) 918K 5) 795K 5 Pressure, [atm] [B] 937K 1060K MCH/air, = Time, [ s] Figure 4.9 HPST pressure-time histories for MCH/air ( =1.0) ignition. Left: near 45atm. Right: near 20atm. Comparisons of an experimental pressure profile with the Ranzi [48] and the Pitz et al. [22] model results are presented in Figure 4.8 for a 46.4 atm test condition. These comparisons serve to confirm that up to the point of ignition, the constant U,V assumption is a good representation of the shock tube behavior (Davidson and Hanson [102]). Because of the nature of the constant U,V assumption, predictions by both mechanisms achieve the same final pressure plateau (at ~ 5 ms). 54

72 Modeled pressure-time histories for MCH/air at 20 atm using the Pitz et al. [22] mechanism (Figure 4.8) do not duplicate the experimental DP/Dt; however, a two-stage pressure increase at low-temperatures (725 K and 775 K) and an approximately linear increase in pressure P 5 at high-temperatures (825 K, and 875 K) are observed. As seen from Figure 4.8, in modeled pressures, typically even with the simple constant U,V assumption, the P 5 rises to about 25 atm (starting from 20atm) before ignition due to preignition chemical heat release alone. The effect of DP/Dt on τ ign may need to be taken into consideration when comparing modeled and experimental results especially at longer ignition delay times [119]. Though developed to deal principally with facility-dependent non-idealities in shock tubes, the CHEMSHOCK [103] program from our laboratory can be applied along with the measured pressure profiles to assess the effect of DP/Dt on the modeled τ ign. A similar approach was used by Pang et al. [141] and they concluded that the use of the experimental pressure trace and the CHEMSHOCK model more accurately modeled the reflected-shock ignition process in hydrogen than the traditional approach using CHEMKIN with a constant U,V constraint. This new approach allows for combined facility-dependent effects and energy release phenomena in the reflected shock environment. It should be noted that CHEMSHOCK assumes stationary, homogeneous conditions within the test gas volume monitored but does not require homogeneity throughout the reflected shock region. The CHEMSHOCK ignition delay times predicted by the Orme et al. [36] mechanism using the measured pressure profile are only 15% higher than the experimentally measured τ ign value at P 5 = atm and T 5 = 912 K (note that this P 5 and T 5 represent an extreme case where τ ign is long). In the current simulations, the pressure was approximated by assuming a linear extrapolation beyond the experimental ignition point (with the same slope, i.e., DP/Dt) until simulation showed ignition. As CHEMSHOCK is a zero-dimensional approach, this latter assumption may force the modeled gases to artificially ignite faster due to the extended increase in pressure (and temperature) beyond the actual experimental ignition point. Since the experimental pressure trace does not differentiate between a rise (i.e., DP/Dt) due to nonideal gasdynamics or to energy release from chemical reaction (somewhere in the reflected shock region), caution should be applied in drawing kinetic conclusions about 55

73 predictability of MCH kinetic mechanisms when used with the CHEMSHOCK modeling approach. Efforts are currently underway in our laboratory to develop and validate more advanced versions (one-dimensional and above) of the CHEMSHOCK program to model shock tube experiments. 4.4 Comparison of Jet Fuel and Surrogate Component Ignition Shown in Figure 4.10 is a comparison of high-pressure shock tube ignition delay times (all from the Stanford HPST) for =1.0 in air (synthetic: 79% N 2 and 21% O 2 ) at 20atm for jet fuel and important jet fuel surrogate components used in the previous chapter. Apart from current MCH data (scaled using τ ign ~ P ), Figure 4.10 includes data for the following fuels: Jet-A from previous chapter (scaled using τ ign ~ P -1 ), iso-octane from Davidson et al. [115] (scaled using τ ign ~ P ) and toluene from Vasu et al. [143] (scaled using τ ign ~ P -1, respectively), n-heptane (high-temperature data Gauthier et al. [39] and NTC data in the range K from current work), and n-dodecane from current chapter (scaled using τ ign ~ P -1 ). In the high-temperature region (above 960K), toluene shows the longest ignition time (aromatics are known to be least reactive), and Jet-A, n-dodecane and MCH have the shortest ignition time. Ignition time results for isooctane and n-heptane, which are also primary reference fuels for gasoline and surrogates, lie in between (note that τ ign results in iso-octane and n-heptane are very close to those of gasoline measured by Gauthier et al. [39] in the HPST). n-dodecane, which is the main n-alkane component of most jet fuel and has physical and chemical properties very similar to that of jet fuels, is very reactive for the conditions shown. It should be noted that the starting temperature for the NTC-type roll-off of MCH is very similar to that of jet fuels above 800K, and among all the single-component fuels considered here, MCH has the closest ignition times to that of Jet-A at high pressures. Large n-alkanes (here n- dodecane), which are main components of jet fuel surrogate mixtures, ignite faster under mild conditions than small or branched chain alkanes in the NTC region and roll-off much earlier and exhibit stronger NTC behavior than Jet fuel (see Figure 4.10). 56

74 Results in Figure 4.10 support conclusions from the previous chapter that in the hightemperature region, using a single-component surrogate (such as MCH above 800K or n- dodecane above 1000K) may be adequate to represent certain combustion characteristics (such as ignition delay time) of jet fuels. However, it may be necessary to use multicomponent surrogates in simple surrogate mixtures for jet fuels that can accurately reproduce ign data in the entire temperature regime (especially in the NTC region below 800K). It should be noted that detailed mechanisms containing a large number of species and reactions (such as in a multi-component surrogate mixture and mechanism) cannot easily be applied to the modeling of combustion behavior of jet fuels in computational fluid dynamics (CFD) calculations. In such cases, modelers can use MCH as a single component surrogate to simulate global combustion kinetic properties of jet fuels Fuel/air, =1.0, 20 atm Ignition Delay Times, [ s] ) Iso-octane 2) Toluene 3) Jet-A 4) n-dodecane 5) MCH 6) n-heptane / T, [1/K] 1 Figure 4.10 High-pressure ignition times data from HPST in Jet-A and major singlecomponent surrogate fuels at similar conditions ( =1.0). HPST data scaled to 20 atm using respective pressure scaling for individual fuels (see text). Solid lines are fit through data at 20 atm: iso-octane and toluene data from Davidson et al. [115] and Vasu et al. [143], respectively; n-dodecane, Jet-A, MCH, and n-heptane are from current work Hydrocarbon ignition is, to a large extent, controlled by the chemistry of the small transient radical pool (H, OH, CH 3 etc.), and in particular, very little or no information is available for these species. At higher pressures, most current mechanisms have been 57

75 validated only against the measured yields of the more stable intermediates or against ignition delay times alone, and small radical species time-history data are needed for complete mechanism validation. Hence, OH concentration time-histories using laser absorption techniques to this end, and the results of these experiments are presented in the next chapter. Table 4.1 Summary of current shock tube experimental results in n-dodecane/air, phi=0.5. n-dodecane/air, phi=0.5, T 1 =100.5C, n-dodecane=0.565 %, O 2 =20.89 %, N 2 =78.55 % T 5 (K) P 5 (atm) τ ign (μs)

76 Table 4.2 Summary of current shock tube experimental results in n-dodecane/air, phi=1.0. n-dodecane/air, phi=1.0, T 1 =103 C, n-dodecane=1.123 %, O 2 =20.77 %, N 2 =78.10 % T 5 (K) P 5 (atm) τ ign (μs)

77 Table 4.3 Summary of current high-pressure ignition time results in MCH/air ( =1.0). Mixture: MCH=1.96%, O 2 =20.60%, N 2 =77.44%. V shock is the incident shock velocity at the endwall. T 5,(K) T 1, (C) V shock,(mm/us) Atten (%/m) T 5, (K) P 5, (atm) ign, (μs)

78 Chapter 5: OH Time-histories During Surrogate Oxidation at High-Pressure This chapter describes high-pressure OH time-histories measured in three singlecomponent surrogates (n-dodecane, MCH, and n-heptane) using a heated HPST. Experimental conditions covered temperatures of 1121 to 1422 K, pressures of atm, and initial fuel concentrations of 500 to 1000 ppm (by volume), and an equivalence ratio of 0.5 with O 2 as the oxidizer and argon as the bath gas. Detailed comparisons of experimental data with predictions of available kinetic mechanisms were made and the procedure was shown to improve predictive qualities. 5.1 Experimental Method All experiments were carried out in the reflected shock region of the HPST. The method and diagnostics are described earlier in a previous chapter. Research grade n- dodecane ( 99%), MCH ( 99.5% pure) and n-heptane ( 99.5% pure) fuel vapors (all supplied by Sigma-Aldrich) and high-purity gases (Praxair O 2, Ar, > %) were prepared similar to that described in previous chapters. The shock tube was heated to 60 C and the fuel/oxidizer mixing facility and connecting lines were kept at a higher temperature (135 C for n-dodecane, and up to 110 C depending on the fuel-loading requirements in the case of MCH and n-heptane) to avoid condensation. The amounts of fuel in mixtures (500 to 1000 ppmv) for the current study were large enough to achieve sufficient absorption signals (and also high signal-to-noise ratios), but small enough to avoid significant energy release during ignition. This ensured uniform temperature and pressure conditions behind the reflected shock wave during reaction. While all data acquired in this study were with the optical access port located 10mm from the endwall, our past experience (Petersen et al. [44]) has shown that there is little sensitivity of the data to small changes in the measurement location. In that study conducted in the HPST (for similar mixtures in methane/o 2 /Ar) no difference was observed between measured peak (time at peak value of X OH for post ignition) for two measurement locations (10 mm and 20 mm from the endwall). A key implication is that 61

79 potential combustion-related gas dynamic effects reported in the previous chapter had minimal impact on our current experiments. There are numerous ways of defining ignition delay time. ign was defined here as the time to reach 50% of the peak X OH, with zero time being defined as the arrival of the reflected shock at the side-wall location. This definition was found to correspond very well with ignition delay defined as the time to the first rise in pressure after arrival of the reflected shock front [144] and has been widely used [42,145]. Additionally, model calculations done by Davidson et al. [42] indicated that this time coincided with the maximum of the CH concentration that occurred during the most rapid formation of radicals in the final stages of ignition for n-alkanes. Example OH absorption traces are shown later in this chapter, however, it is to be noted that the spike in the OH trace at time zero corresponds to arrival of the reflected front at the diagnostic location. The diagnostic beam is temporarily steered (due to the density gradient across the reflected shock wave) off the detector surface resulting in the observed schlieren spike. The arrival of the reflected shockwave at the sidewall location (and therefore time zero) was determined both from the step rise in pressure signals and from the OH traces as the midpoint of this above-mentioned spike. 5.2 Kinetic Modeling Details In comparing OH experimental data with predictions of available mechanisms, the calculations were done for homogeneous, adiabatic conditions behind reflected shock waves, with the common constant-internal-energy, constant-volume constraint (constant U,V) using CHEMKIN [101]. For short ignition delay times (typically less than 1-2 ms) in dilute fuel-oxidizer mixtures, the constant U, V constraint is a good assumption for the purpose of ignition delay time calculations [102,103]. In the modeling, an identical ign definition to experiments was used, i.e., the time to reach 50% of the peak X OH. We have compared our OH species measurements with the predictions of three MCH mechanisms, 11 n-heptane mechanisms, and 4 n-dodecane mechanisms as listed in Table 5.1. The mechanisms discussed here have been developed and optimized to address particular needs and applications, and have not, in particular, been optimized to accurately simulate OH time-histories. However, the importance of the small radical pool 62

80 (H, O, HO 2, OH, and CH 3 ) to all ignition processes implies that these mechanisms should realistically be expected to model OH profiles accurately. It is important to look at the details of kinetic mechanisms used for comparison in order to get a better sense of the chemistry and the similarities and differences in predictions. The MCH mechanism of Orme et al. [36] was built on the reaction scheme developed by Laskin et al. [56], which described the oxidation of 1,3-butadiene, although the H 2 /O 2 submechanism was based on that of Ó Conaire et al. [146]. The Orme et al. mechanism was validated against shock tube ignition delays and flow reactor experiments. The Pitz et al. [22] mechanism contained C 1 -C 6 chemistry of Curran et al. [113] with submechanisms for toluene, benzene, and cyclo-pentadiene developed by the same authors. It is important to note that the comprehensive detailed kinetic mechanism for MCH oxidation from Pitz et al., which adapted the high-temperature MCH reactions from Orme et al. [36], was validated using RCM ignition delay time measurements but has not previously been validated for temperatures above 1100 K. The Ranzi [48] (semi-lumped) JP-8 surrogate mechanism contains n-dodecane, MCH and n-heptane. The You et al. [118] n-dodecane mechanism was a recently developed kinetic mechanism for the combustion of n-alkanes up to n-dodecane, where n-heptane is an integral sub-mechanism component and is currently part of the JetSurF 1.0 [46] surrogate mechanism. The Montgomery et al. [117] quasi-steady-state reduced mechanism, created using an automatic genetic optimization scheme, is based on the parent Violi et al. [5] mechanism which has the same origin as the current Ranzi [48] mechanism. The semi-detailed JP-8 mechanism of Zhang et al. [19] was not optimized for low temperature modeling. The high-temperature Chaos et al. [52] n-heptane mechanism, which was developed based on the baseline H 2 /O 2 and C 1 -C 4 chemistry of Li et al. [147] and Zhao et al. [148], respectively, was validated against a variety of experiments including shock tube ignition delays, premixed flame speeds, and species measurements from a variable pressure flow reactor (VPFR) and a jet-stirred reactor (JSR). The large n-heptane mechanism of Curran et al. [149], also known as LLNL (detailed), was validated against shock tube, RCM, and flow and stirred reactor data. The Seiser et al. [150] mechanism, also referred to as LLNL (reduced), was reduced from [149] using data for n-heptane extinction and autoignition in 63

81 counterflow configuration. The n-heptane mechanism of Tsang [151] (denoted as the NIST mechanism) was based on a variety of other mechanisms and was validated against burning velocity, ignition delays and low-pressure OH time-histories (by Davidson et al. [42]) in a shock tube. The Golovitchev [152] mechanism was validated against a variety of data including shock tube ign. The highly reduced Patel et al. [153] mechanism used [152] as a starting point and was validated using single-cylinder engine data. The San Diego [154] n- heptane mechanism was built by adding lumped n-heptane chemistry to the widely validated C 3 mechanism developed by Petrova and Williams [155]. The Biet et al. [156] n-heptane mechanism (not validated at high-temperature conditions) has been generated using the EXGAS software, a computer package developed to perform the automatic generation of detailed kinetic models for the gas-phase oxidation and combustion of alkanes. The Gokulakrishnan et al. [157] kerosene mechanism includes the n-heptane skeletal mechanism derived from [149] and was validated for ignition delay times. 5.3 OH Time-Histories Results All new experimental data for X OH obtained in this study are summarized in Table 5.2 and Table n-heptane Results The Measured OH profiles during n-heptane oxidation are shown in Figure 5.1. All data show approximately the same behavior, i.e., there is no significant OH formation at early times (during initial fuel decomposition), and during ignition the OH concentration rapidly increases, reaches a peak value and slowly falls off. This behavior of OH is different from the behavior observed in low-pressure experiments, where at early times an OH plateau (simultaneous with the initial fuel decomposition of the fuel molecule) is observed, after which OH rises to a higher post-ignition plateau in most fuels [42,68,145]. A low temperature result at 1121K, a case where n-heptane did not ignite during the time interval of interest in the current study (< 2 ms), is also shown, indicating that there is no pre-ignition formation of OH even at longer times for the current experiments. As expected, peak X OH values decrease and ign increases as temperature decreases. 64

82 The ability of the n-heptane mechanisms investigated to model the current n-heptane OH profile data is shown in Figure 5.2 for the case of 1271K. Evidently, there is significant difference also in the shape of OH profiles predicted by various mechanisms. Wide variations exist in predicted OH profiles for the conditions shown at 1271K. Notably, the shape of OH profiles of Patel et al. [153], Ranzi et al. [48], SanDiego [154], Chaos et al. [52], Biet et al. [156], and You et al. [118] mechanisms are very similar to that of the HPST data. Profiles predicted by Tsang [151] and both the LLNL mechanisms [149,150] are similar (there is little difference between predictions of the detailed and reduced LLNL mechanisms) while the Golovichev [152] mechanism predicts a very different shape compared to the data. Note that because the model predictions are performed assuming constant U,V constraints, the final plateau value (after decay) obtained is same for every mechanism X OH, [ppm] Time, [ s] 4 Figure 5.1 High-pressure OH absorption data for n-heptane; initial X fuel =1000ppm, X O2 =0.022, X Ar =0.977, =0.5. Data 1: 1271K, 15atm; data 2: 1236K, 15.28atm; data 3: 1230K, 15.81atm; data 4: 1121K, 14.1atm. The data show no significant early formation of OH, and all the n-heptane mechanisms predict this correctly except for You et al. [118] where there is an approximately 10ppm early OH plateau at a temperature of 1271K, and pressure of 15 atm, for an initial X n-heptane =1000ppm. Only the You et al. mechanism predicts shorter ignition delay times than experiment, and this fact is consistent with the predicted formation of early-time OH concentrations. Some mechanisms (Ranzi, Patel et al., Biet et al., Chaos et al.) predict the peak X OH values within 20%, while over-predicting ign by a significant amount. Patel et al. predicts a very long ignition time, which is understandable 65

83 considering that it is a highly reduced mechanism developed for fast multi-dimensional diesel HCCI engine simulations and may not be suitable for modeling OH radical concentrations. The San Diego [154] mechanism gives the closest agreement by predicting all parameters within 20% for the current experiments. Both the LLNL mechanisms [149,150] under-predict peak X OH and over-predict ign by 50 %. The Horning [43] empirical correlation for n-heptane from our lab, which was developed by regression analysis of ignition time measurements of n-heptane conducted over a range of temperature, pressure (1-6atm) and mixture composition, predicts ign within 10% of the current data X OH, [ppm] X OH, [ppm] , Time, [ s] Time, [ s] Figure 5.2 High-pressure OH absorption profile data and modeling predictions for n- heptane; initial X fuel =1000ppm, X O2 =0.022, X Ar =0.977, =0.5, 1271K, 15atm. 1: Current experiment, 2: Curran et al. [149], 3: Seiser et al. [150], 4: Tsang [151], 5: Patel et al. [153], 6: Ranzi et al. [48], 7: SanDiego [154], 8: Golovichev [152], 9: Gokulakrishnan et al. [157], 10: Chaos et al. [52], 11: Biet et al. [156], 12: You et al. [118] MCH Results The Example measured OH profiles during oxidation of MCH are shown in Figure 5.3 for two initial fuel concentrations at different temperatures. All data show approximately the same behavior and the characteristics are similar to those of n-heptane. High reproducibility of the current experiments (as a result of accurate mixture concentration and shock-speed determinations, and consistency of the diagnostic technique) and sensitivity of the measured OH profiles to temperature are evident from Figure 5.3. Figure 5.4 shows a comparison of the OH profile data in MCH at 1262 K with the 66

84 predictions of all three MCH models. None of the mechanisms predict OH production consistent with the early-time data. The semi-lumped mechanism of Ranzi [48] recovers both the shape and peak value of OH, while the Orme et al. [36] mechanism matches only the decay rate. The Pitz et al. [22] mechanism predicts a slower rise during ignition and a near flat peak (also much smaller) compared to the sharp peak observed during experiments. The measured peak X OH follows an approximately linear dependence on temperature. The OH peak concentrations for the 1000 ppm mixtures are observed to be 1.7 times the OH peak concentrations in the 750 ppm mixtures. All three mechanisms recover the temperature dependence of the observed peak OH. The Ranzi et al. mechanism is in very good agreement with the values from the experiment, while the other two mechanisms under-predict the peak value at all conditions investigated. 200 A 1) 1285 K 2) 1269 K 3) 1213 K B 1) 1304 K X OH, [ppm] X OH, [ppm] ) 1303 K ) 1266 K 4) 1205K Time, [ s] Time, [ s] Figure 5.3 (A): OH absorption data for MCH; initial X MCH =1000 ppm, X O2 =0.021, X Ar =0.978, =0.5. Data 1: 1285 K, atm; data 2: 1269 K, 15.8 atm; data 3: 1213 K, atm; data 4: 1205 K, atm. (B): OH absorption data for MCH; initial X MCH =750 ppm, X O2 = , X Ar =0.9835, =0.5. Data 1: 1304 K, atm; data 2: 1303K, atm; data 3: 1266 K, atm. Comparison of high-temperature, low-concentration ignition delay time measurements and modeling predictions show that at fixed equivalence ratios, ignition delay times increase with decreases in either temperature or initial fuel mole concentration. In general, all three mechanisms do a reasonable job of predicting the ignition delay times, with the Ranzi and the Orme et al. mechanisms giving closer agreement with our data than the 67

85 Pitz et al. mechanism, for both initial fuel concentrations. The following activation energy for ignition predicted by all three mechanisms is slightly higher than the value (38 kcal/mol) obtained during current experiments: kcal/mol by Ranzi; kcal/mol by Pitz et al.; and kcal/mol by Orme et al. This experimental activation energy for ignition of MCH is lower than those observed in most of the n-alkanes up to C 10 [43]. X OH, [ppm] HPST Ranzi et al Orme et al. Pitz et al Time, [ s] Figure 5.4 High-pressure OH absorption data and modeling predictions (using Ranzi [48], Orme et al. [36], Pitz et al. [22] mechanisms) for MCH/O 2 /Ar. Initial X MCH =1000 ppm, X O2 =0.021, X Ar =0.978, 1262K, atm, = n-dodecane Results The measured OH profiles for various temperatures for an initial X fuel of 1000ppm are plotted in Figure 5.5. At the highest temperature (1422K), there is clear evidence of early-time radical chemistry (see inset, Figure 5.5), i.e., there is the rapid formation of OH and attainment of an intermediate plateau level (X OH ~20ppm) simultaneous with the initial decomposition of the fuel molecule. All four mechanisms predict this behavior at 1422 K, but the intermediate plateau level is different for each mechanism: Ranzi [48](16ppm), Montgomery et al. [117] (18ppm), Zhang et al. [19] (10ppm) and You et al. [118] (57ppm). At lower temperatures, there is no measurable OH at early time (nor is a significant OH plateau seen in any of the models.). Figure 5.5 also shows a comparison of the 1217K data with all four model prediction. The You et al. and the Montgomery et al. mechanisms recover the shape of the OH peak at ignition, while there is a sharp overshoot evident in the Ranzi and Zhang et al. predictions. 68

86 X OH, [ppm] XOH, [ppm] Time, s] 5 X OH, [ppm] ppm n-dodecane,1217 K, atm, =0.5 You et al. Ranzi et al. 600 Zhang et al. Montgomery et al. Current Study Time, s] Time, s] Figure 5.5 High-pressure OH absorption data: n-dodecane; initial X dodecane =1000ppm, O 2, Ar; =0.5. Left: Data 1: 1422K, 15.5atm; data 2: 1230K, 16.73atm; data 3: 1217K, 16.07atm; data 4: 1196K, 15.77atm; data 5: 1158K, 15.19atm. Right: Comparison of measured and modeled OH time-histories at 1217 K, 16.1 atm. Figure 5.6 shows a comparison of the measured peak OH concentration and ignition delay times predicted by different mechanisms. The measured peak X OH follows an approximately linear dependence on temperature and an approximately linear variation on initial fuel concentration (X fuel ) as well. The measured OH peak concentrations for the 1000 ppm mixtures are 1.7 times the OH peak concentrations observed in the 515 ppm mixtures (not shown here). All four mechanisms recover the temperature dependence of the observed peak OH. Ignition delay times also can be determined from the hightemperature, low-concentration OH concentration time-histories. Figure 10 shows a comparison of τ ign (the time to reach 50% of the peak post-ignition X OH ), with the predictions of the four models for the 1000 ppm n-dodecane, =1 data. τ ign values depend strongly on the initial X fuel concentration, i.e., τ ign for initial X fuel =515 ppm mixtures are 2.7 times those for initial X fuel =1000 ppm mixtures. The You et al. model simulations gives the closest agreement (within 10%), and the Zhang et al. mechanism gives the poorest. The activation energies predicted by the Montgomery et al. gives slightly better (within 40%) agreement to the data than the Ranzi mechanism predictions. 69

87 Peak OH Mole Fraction [ppm] 1333K K 1000ppm n-dodecane/o 2 /Ar 16 atm, = K Ranzi et al. You et al. Zhang et al. Montgomery et al. Current Study 1111K /T, [1/K] Ignition Delay Times, [us] K 1250K Current Study You et al. Zhang et al. Montgomery et al. Ranzi et al. Horning et al. 1000ppm n-dodecane/o 2 /Ar = K /T, [1/K] Figure 5.6 High-pressure OH data variation with temperature. Initial reflected shock conditions: 16 atm, 1000 ppm n-dodecane/o 2 /Ar, = 0.5. Left: peak X OH vs T 5. Solid line is a linear fit through data. Right: High-temperature, low-concentration ignition delay times in n-dodecane. Solid black line is least-squares fit through data. Figure 5.6 also shows the predictions of the ignition time correlation of Horning et al. [43], developed at low pressures for a variety of n-alkanes. This correlation was developed for n-alkanes from n-propane to n-decane and accounts for the fuel size through the number of carbon atoms in the molecule. While over-predicting the n- dodecane ignition delay time at high pressures, it does accurately capture the activation energy of the current data set. The activation energy for current n-dodecane data is 46 kcal/mol, which is very close to that obtained by Horning et al. [43] (46.5 kcal/mol) and You et al. [118] (45.5 kcal/mol). Montgomery et al. [117] (43.5kcal/mol), Ranzi [48] (42.2 kcal/mol) and Zhang et al.[19] (52.9 kcal/mol) do not agree as well. 5.4 Discussion In order to understand fuel oxidation chemistry and to identify key reactions which affect OH formation with a view to improve the surrogate modeling predictive capabilities, kinetic analysis was performed. Rate of production (ROP), reaction pathway, and OH sensitivity analysis were performed and are discussed. The OH sensitivity coefficient is defined as: S(X OH, k i, t)= {dx OH (t)/dk i }{k i /X OH (t)}. 70

88 5.4.1 n-heptane Kinetic Analysis The mechanism of Chaos et al. [52] was chosen to perform integrated reaction flow analysis because of its compactness, thorough and extensive validation against a large experimental data set covering a broad range of applications, and its ability to predict current values of peak X OH. The results of the flow analysis are summarized in Figure 5.7, which identifies dominant reaction pathways for n-heptane oxidation (1275K, 16atm). Approximately 80% of n-c 7 H 16 is removed by H-abstraction reactions, primarily by reaction with OH. The heptyl (C 7 H 15 ) radicals which are formed further undergo C-C bond cleavage to form four main product channels. The decomposition channels of C 7 H 15 are 1-butene (1-C 4 H 8 ) and n-propyl radical (n-c 3 H 7 ) (48%), 1-pentene (1-C 5 H 10 ) and ethyl radical (C 2 H 5 ) (35%), and two minor channels (~8% each) producing C 3 H 6, C 2 H 4, C 5 H 11 (pentyl radical), and p-c 4 H 9 (1-butyl radical). Continued decomposition of these species forms the relatively stable intermediates CH 3, C 3 H 6, and C 2 H 4. Figure 5.7 Major oxidation pathways prediction using the Chaos et al. [52] mechanism integrated ROP approach. X heptane =1000ppm, X O2 =0.022, balance=ar. 1275K, 16atm, =0.5. Details of molecular structures and pathways can be found in [52]. Based on the results of sensitivity analyses and published uncertainties of several of these reactions, likely reaction candidates for mechanism adjustment can be selected. As expected, sensitivity analysis for OH (shown in Figure 5.8) using the Chaos et al. [52] 71

89 mechanism reveals that the OH concentration both at early times and during ignition is most sensitive to the branching reaction:h+o 2 =O+OH. However, this reaction has been extensively studied over the years, and recent publications place an uncertainty of less than 10% on this rate coefficient [158]. In addition, OH time histories are strongly impacted by the following reactions: CH 3 +HO 2 =CH 3 O+OH, and C 2 H 4 +OH=C 2 H 3 + H 2 O. Note that CH 3 +HO 2 reaction has two product channels and they are in direct competition with each other (see Figure 5.8).The substantial amount of reactive H radicals produced from CH 3 O through the subsequent decomposition reaction CH 3 O+M=CH 2 O+H+M explains the very strong positive sensitivity of this channel. However, even though CH 3 +HO 2 =CH 3 O+OH has a strong influence on τ ign, there exists no direct measurement of this reaction and differences exist in the rate used by the modeling community. Based on the value (1.81x10 13 cm 3 /mole/s) suggested by Colket et al. [159], the Baulch et al. [160] review suggested an uncertainty factor of 10 for CH 3 +HO 2 =CH 3 O+OH and another review by Tsang and Hampson [127] estimated a temperature-independent value of 1.99x10 13 cm 3 /mole/s, with an uncertainty factor of Normalized OH Sensitivity A) H+O2<=>O+OH B) HO2+OH<=>H2O+O2 C) OH+C2H4<=>C2H3+H2O D) HO2+CH3<=>O2+CH4 E) HO2+CH3<=>OH+CH3O F) C3H6+OH<=>aC3H5+H2O A E C D, F B Time, [ s] Figure 5.8 OH sensitivity for n-heptane oxidation using the Chaos et al. [52] mechanism. X heptane =1000ppm, X O2 =0.022, balance=ar. 1271K, 15atm, =0.5. The CH 3 +HO 2 =CH 3 O+OH values (in units of cm 3 /mole/s) used by n-heptane mechanisms are 1.8 in Biet et al. [156], 1.48 in Chaos et al. [52], and 1.1 in Seiser et al. [150], while the most popular and extensively validated, GRI 3.0 [161] optimized methane mechanism uses The recent OH time-history measurements during iso- 72

90 octane oxidation in a RCM for pressures between atm by He et al. [162] obtained good agreement between data and predictions by a detailed iso-octane mechanism (Curran et al. [113]) when a value of 7.7x10 13 cm 3 /mole/s was used. We find that using a value of 6.8x10 13 cm 3 /mole/s (the theoretical rate constant of CH 3 +HO 2 =CH 3 O+OH calculated by Zhu and Lin [163] is only 50% lower than this value) with the Chaos et al. [52] mechanism significantly improves the ign predictive capability for the current n- heptane measurement at 1271K (see Figure 5.9).The values of the peak X OH changes only minimally by this alteration. Our observations highlight the necessity of further, more direct study of this reaction in order to obtain more accurate values and hence reliable predictions for ign. At long times (after about 690 s as in Figure 5.9), there is disagreement between modified Chaos et al. [52] mechanism predictions and data, but these differences may be due to post-ignition waves or flow disturbances inside the shock tube, which cannot be simulated using the current modeling approach, i.e., using constant U,V CHEMKIN [101]. 250 Chaos et al. mechanism modified unmodified 200 X OH, [ppm] current data Time, [ s] Figure 5.9 Influence of higher rate for CH 3 +HO 2 =CH 3 O+OH, new rate= 6.8 x10 13 cm 3 /mole/s, on the OH predictions by Chaos et al. [52] for n-heptane oxidation. X heptane =1000ppm, X O2 =0.022, balance=ar, =0.5, 1271K, 15atm. The OH+C 2 H 4 reaction influences OH concentrations not only in n-heptane oxidation but in MCH and n-dodecane oxidation as well. This reaction accounts for 28% (Chaos et al. [52]), 37% (Orme et al. [36]), and 50% (You et al. [118]) of the C 2 H 4 removal during 73

91 n-heptane, MCH, and n-dodecane oxidation, respectively, at 1275 K and 16 atm. Uncertainty factors of 5.0 and 3.16 are listed for this reaction from reviews conducted by Tsang and Hampson [127] and Baulch et al. [160], respectively. Increasing the rate of this reaction, which produces vinyl radicals, also improves (i.e., shortens ign without affecting peak X OH ) the predictions by the Chaos et al. [52] n-heptane mechanism (not shown here). This reaction was measured in this work and the results are presented in a later chapter MCH Kinetic Analysis The Orme et al. [36] mechanism was used to perform integrated flow analysis for MCH oxidation at 1275 K and 16 atm (similar conditions to those presented in the case of n-heptane in Figure 5.7), and the simplified oxidation pathways are presented in Figure For clarity, chemistry below the C 2 level is not shown, and symbols are used to represent some complex species without affecting the purpose of the current discussion. The Orme et al. mechanism was chosen due to its simplicity and the relatively encouraging results (peak shape) in Figure 5.4. As shown in Figure 5.10, the importance of 1,3-butadiene chemistry in MCH oxidation is clearly evident as nearly all of MCH decomposes through 1,3-butadiene. Also, typical modeled species concentrations during MCH oxidation using the Orme et al. mechanism indicate that 1,3-butadiene is the second largest intermediate species following ethylene. 1,3-butadiene reacts with H atoms (40%) to form ethylene and vinyl (C 2 H 3 ) radicals, and about 30% of 1,3-butadiene goes to form allyl radicals. It is important to note that allylic radicals determine the concentrations of many important aromatic precursors in oxidative systems. Here, allyl forms C 3 H 6 through +H (+M) reactions, 1-butene through +CH 3 (+M) reactions, and vinyl plus formaldehyde through reacting with HO 2. These analyses highlight the importance of 1,3-butadiene, HO 2 and C 2 H 4 chemistry during MCH oxidation. Additionally, note that two of the major stable intermediates during MCH oxidation, C 2 H 4 and C 3 H 6, are consumed through reactions with OH radicals, just as in the case of n-heptane oxidation presented earlier, i.e., 40% of C 3 H 6 and 37% of C 2 H 4. Ethylene is also consumed (39%) by its reaction with O atom to produce methyl and formyl (HCO) radicals. 74

92 Figure 5.10 Major MCH oxidation pathways using the Orme et al. [36] mechanism (integrated ROP), 1275 K, 16 atm, 1000 ppm MCH/O 2 /Ar, = 0.5. Refer to Orme et al. [36] for IUPAC nomenclature of species and for detailed molecular structures and pathways. Normalized OH sensitivity analysis for MCH oxidation is shown in Figure 5.11 using the Orme et al. [36] mechanism for the OH experimental conditions shown in Figure 5.4. Besides the reactions mentioned in the case of n-heptane oxidation, OH sensitivity is affected by the MCH decomposition through the cyclo-hexyl radical channel:mch=cyclo-hexyl + methyl. Immediately after the arrival of the reflected shock (at very early times), the influence of this reaction (labeled A in Figure 5.11) on the OH concentrations is higher than that of H+O 2 =O+OH. The influence of the decomposition channel decreases subsequently and the methyl radical recombination reaction becomes dominant along with H+O 2 =O+OH (see Figure 5.11). Hence, modeling the early time MCH decomposition could be improved by adjusting the above-mentioned rates within their uncertainty bounds. OH scavenging reaction by HO 2 radical has the strongest negative sensitivity values during ignition in both the cases of n-heptane and MCH. However, adjustments to this reaction (HO 2 +OH=H 2 O+O 2 ) were not studied here, because this reaction is well-known, and very recent measurements of this reaction [158] yield an estimated uncertainty of less than 40%. 75

93 The most important 1,3-butadiene reaction occurring in the MCH sensitivity analysis of OH near ignition is the reaction (which is also the main 1,3-butadiene decomposition channel):1,3-butadiene+h=c 2 H 4 +C 2 H 3. Tsang and Hampson [127] estimated an uncertainty factor of 5.0 for the reverse reaction. There exist no direct measurements and the rate used by Orme et al. [36] is the same as that used in the original Laskin et al. [56] 1,3-butadiene mechanism. However, Libby et al. [68] suggested a 3-times lower rate for this reaction, than that estimated by Laskin et al., based on fitting their low-pressure OH time-history measurements with the predictions made by the Laskin et al. 1,3-butadiene mechanism. Normalized OH Sensitivity A) Cyclo-hexyl+CH3=MCH B) H+O2=O+OH C) HO2+OH=H2O+O2 D) CH3+HO2=CH3O+OH -40 E) C2H4+OH=C2H3+H2O F) 1,3-butadiene+H=C2H4+C2H3 C Time, [ s] a B D A,F E Normalized OH Sensitivity A B G A) Cyclo-hexyl+CH3=MCH B) H+O2=O+OH G) CH3+CH3(+M)=C2H6(+M) b Time, [ s] Figure 5.11 OH Sensitivity for MCH oxidation using the Orme et al. [36] mechanism. X MCH =1000 ppm, O 2 ( =0.5), balance=ar K, atm. (a): during ignition, (b): at early times. Because of the importance of 1,3-butadiene in MCH oxidation systems, a comparison of the current MCH mechanisms to predict OH time-histories during 1,3-butadiene (with Libby et al. [68] data) oxidation was performed. It was found that none of the MCH mechanisms predict ign or the early time OH plateau before ignition, and only the Ranzi et al. [48] predicted the plateau of OH value after ignition. The Ranzi et al. JP-8 mechanism does not distinguish between isomers of 1,3-butadiene and all butadiene isomers along with butyne are treated as a single lumped species. The Orme et al. [36] ignition time predictions are approximately 10% shorter than the Laskin et al. [56] predictions, which are about 36% shorter than those of the Libby et al. [68] data. In order to improve the predictive capabilities of Laskin et al. [56] for hightemperature 1,3-butadiene oxidation, Libby et al. [68] suggested the following 3 76

94 recommendations: 1) the GRI 3.0 [161] recommended rate coefficient for H+O 2 =O+OH, 2) a lower heat of formation of OH, Δ f H 298 (OH)=8.92 kcal/mole, as measured recently by Herbon [97] and Ruscic et al. [164], and 3) a 3-times lower rate for 1,3- butadiene+h=c 2 H 4 +C 2 H 3. The modified OH profile predictions (not shown here, see Libby et al. [68]), using the Laskin et al. [56] 1,3-butadiene mechanism, after implementing these 3 recommendations, very closely matched with the Libby et al. data. The effect of the second recommendation is that predictions of the plateau values were improved, i.e., the OH plateau of Laskin et al. matched that of Libby et al. The Orme et al. [36] mechanism already uses this latest value for heat of formation of OH and the plateau predictions are better for Orme et al. compared to Laskin et al. It must be necessary to include the Libby et al. recommendations 1 and 3 in the Orme et al. mechanism as well. ign and early time OH plateau predictions by Orme et al. during 1,3-butadiene oxidation are improved, but prediction of the shape of the OH profile is destroyed, which showed early and much slower rise to the experimental plateau value after ignition. Note that modifications were applied in an incremental fashion, i.e., each modification is applied step by step to see the effect of each recommendation. Overall, it was found that adding MCH chemistry to the Laskin et al. 1,3-butadiene oxidation reactions without validating the modified mechanism against 1,3-butadiene experiments, i.e., the procedure used to develop the Orme et al. MCH mechanism, affects the predictive capability of the MCH mechanism to model 1,3-butadiene oxidation. When the Libby et al. [68] recommendations 1 and 3 were included in the Orme et al. [36] mechanism in an incremental fashion, the ign predictions for MCH oxidation are considerably longer than the unmodified Orme et al. predictions (see Figure 5.12). Also, the shape of the OH profile shows a slower rise to the peak X OH and a lower peak X OH, which is now much below the experimental peak value. However, if we include the modifications from Figure 5.9 that were applied in the case of n-heptane oxidation, i.e., the new 6.8x10 13 cm 3 /mole/s value for CH 3 +HO 2 =CH 3 O+OH reaction, ign predictions are now within a few percent of the experimental value. It is clear that, just as in the case of n-heptane, using the new recommendations affects the peak X OH only minimally; however predictions of the shape of the rise of OH are slightly distorted due to this procedure. Finally, the observed effect of increasing the rate of OH+ethylene is similar to 77

95 that observed by increasing CH 3 +HO 2 =CH 3 O+OH values for both MCH and n-heptane. The reasons for the remaining discrepancies in the modeled and measured peak OH concentration are not clear. 200 Modifications to Orme et al. A) Current data C) Libby et al. (1) B) Orme et al D) Libby et al. (1+3) (unmodified) E) Libby et al. (1+3)+ 150 new k 2 value X OH, [ppm] 100 A E 50 B C D Time, [ s] Figure 5.12 Effect of modifications to the Orme et al. [36] mechanism predictions for MCH oxidation (combining the Libby et al. [68] recommendations (1 and 3) and the 6.788x10 13 cm 3 /mole/s value for CH 3 +HO 2 =CH 3 O+OH reaction). Initial X MCH =1000ppm, O 2 ( =0.5), balance=ar. 1262K, 15.45atm. See text for details of the Libby et al. [68] recommendations. The reason for different OH profile shapes and significantly different peak OH concentrations (see Figure 5.4) predicted by the Ranzi [48] and Pitz et al. [22] mechanisms was investigated. For these two mechanisms, the ROP results (for conditions in Figure 5.4) indicate that the most influential reactions that produce OH are similar and the reaction CO+OH=CO 2 +H is the significant OH scavenging reaction. Also, ROP shows that the main OH precursors or scavengers are H, O, HO 2 and CO for both mechanisms. At this condition (as in Figure 5.4), both these mechanisms predict nearly the same maximum amounts of CO (~ 5000ppm) and HO 2 (~ 30ppm), however, the maximum H, O, and OH predicted by the Ranzi et al. mechanism (note from Figure 5.2 and Figure 5.4 that the maximum OH predictions by the Ranzi mechanism are very close to the current experimental results in n-heptane and MCH) are approximately 10, 6 and 3.5 times, respectively, of those predicted by the Pitz et al. mechanism. Naturally, this would suggest that either the Ranzi mechanism uses more accurate rate values for 78

96 important OH reactions (all involving small species) or the Pitz et al. mechanism assumptions regarding larger hydrocarbon chemistry (involving these small radicals) need to be re-examined. However, replacing the Pitz et al. values of several of the important OH reactions (i.e., H+O 2 =O+OH, CO+OH=CO 2 +H, H+HO 2 =OH+OH, and O+H 2 O=OH+OH) with those of the Ranzi mechanism results in only negligible improvements in the Pitz et al. mechanism predictions, which are not sufficient (also supported by our analysis below) to explain the large differences in OH predictions seen in Figure n-dodecane Kinetic Analysis The You et al. [118] mechanism was used to perform integrated flux analysis for n- dodecane oxidation at 1275K and 16atm. The n-dodecane oxidation pathways according to the You et al. mechanism are shown in Figure Rates of production analyses at 1275K and 16atm indicate that H abstraction (by reaction of the fuel with H-atom and to a lesser extent OH) dominates the initial removal of n-dodecane and forms 6 dodecyl radical isomers. The alkyl radicals formed from n-dodecane subsequently undergo reactions producing various lower alkyl and alkenyl radicals, and alkenes, where β- scission dominates the later decomposition channels, forming mainly ethylene and propene along the way. Ethylene and propene hence are two of the most stable intermediate species during n-dodecane oxidation at these conditions (the rapid removal of ethylene coincides with the sharp rise in OH during ignition). The C 2 H 4 +OH=C 2 H 3 +H 2 O, accounts for 50% of ethylene removal under these conditions. These conditions are similar to those presented in the case of n-heptane and MCH in Figure 5.7 and Figure Other n-dodecane kinetic mechanisms are either very large, lumped or do not predict the X OH peak shape profiles accurately well (Figure 5.5). The You et al. [118] mechanism reasonably predicts τ ign, τ peak, and peak X OH. You et al. state that kinetic parameters of their mechanism for higher order hydrocarbon chemistry (greater than C 4 ) were derived from the analogous C 4 reactions from their base model without ad hoc parameter tuning. However (drawing from the case of MCH presented 79

97 earlier), accurately predicting the entire OH time-histories may require more detailed and rigorous treatment of the n-dodecane decomposition process to C 4 or lower hydrocarbons. We have not adjusted the You et al. [118] mechanism or the included rate constants, but we can comment on the reactions in these mechanisms most deserving of further scrutiny. An OH sensitivity analysis was performed using the You et al. [118] mechanism for the highest temperature (1422K) case. At the earliest times when n-dodecane begins to decompose (within 7-10 μs), OH production is most sensitive to the rate of the reaction H+O 2 =O+OH. However, the intermediate OH plateau level seen at this temperature is maintained by a balance between the production of OH through H+O 2 =O+OH and HO 2 +H=OH+OH, and removal of OH through C 2 H 4 +OH=C 2 H 3 +H 2 O and CH 2 O+OH=HCO+H 2 O. The intermediate plateau level and subsequent ignition delay time are thus strongly sensitive to decomposition pathways that form C 2 H 4 and CH 2 O. Sensitivity analyses conducted at 1275K, 16atm is shown in Figure 5.14, which clearly indicates the importance of reactions mentioned earlier (in the case of n-heptane and MCH). Current results for MCH and n-heptane showed that it is possible to achieve excellent agreement between measurements and data by following the procedure outlined in this section. Improved understanding of intermediate chemistry, especially that of CH 3, HO 2 and alkenes (ethylene and propene) for all three fuels and 1,3-butadiene for MCH, is critically needed to better understand the overall fuel oxidation and to improve predictions at high-temperatures and pressures. Such work calls for direct and accurate measurements of the rate constants of key major reactions and these results are presented next. 80

98 Figure 5.13 n-dodecane oxidation pathways using the You et al. [118] mechanism. X NC12H26 =1000ppm, O 2, =0.5, balance=ar, 1275K, 16atm. Normalized OH Sensitivity ) H+O2=O+OH 2) OH+HO2=H2O+O2 3) CH3+HO2=CH3O+OH 4) C2H3+O2=CH2CHO+O 5) C2H4+OH=C2H3+H2O Time, [ s] Figure 5.14 Normalized OH sensitivity results for n-dodecane oxidation using the You et al. [118] mechanism. X NC12H26 =1000ppm, =0.5, balance=ar. 1275K, 16atm. 81