SHOCK TUBE STUDIES OF KINETICS OF CONVENTIONAL AND ALTERNATIVE ENGINE FUELS

Transcription

1 SHOCK TUBE STUDIES OF KINETICS OF CONVENTIONAL AND ALTERNATIVE ENGINE FUELS A DISSERTATION SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Yangye Zhu November 2016

2 2016 by Yangye Zhu. 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. Hai Wang Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost for 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 iv

5 Abstract Conventional fossil-based hydrocarbon fuels account for over 80% of the primary energy consumed in the world. However, the current climate and energy crises make us vigilant in seeking alternative fuel sources and mindful of improving the performance of engines. In these efforts, chemical kinetics have been essential and critical for better engine performance and reducing pollutant emissions. Yet there are major gaps in our knowledge of the high temperature chemistry of real engine fuels, and hence there is a critical need for experimental kinetic databases that can be used for the validation and refinement of engine fuel combustion models. To fill this need, experiments were performed using shock tube and laser absorption methods to investigate the oxidation and pyrolysis systems of conventional and alternative fuels, and fuel components, under engine-relevant conditions. Ignition delays were measured during fuel oxidation and several stable intermediates were measured during fuel pyrolysis. Ignition delay times were measured in the gas-phase for several engine fuels in various oxidizers behind reflected shock waves in a heated high-pressure shock tube. Initial reflected shock conditions were as follows: temperatures of K, pressures of 2-60 atm, equivalence ratios ( ) of 0.2-2, and oxygen concentrations of 4% and 21% with argon and 21% in synthetic air. Ignition delay times were measured using sidewall pressure, OH* emission at 306 nm as well as 3.39 µm mid-infrared He-Ne laser absorbance. The new experimental results were modeled using several kinetic mechanisms with the most current jet fuel surrogate mixtures as well as with a new hybrid model currently in development at Stanford. Cyclo-alkanes are one of the most important chemical classes found in gasoline fuels and diesel fuels. Ignition delay time experiments were conducted during high-pressure oxidation of one representative component for cyclo-alkanes in jet fuel and diesel fuel surrogates, i.e., decahydronaphthalene (decalin). Ignition delay measurements were conducted for gas-phase decalin/air mixtures over temperatures of K, pressures of atm, and equivalence ratios of 0.5, 1.0, and 2.0. Negative-temperature-coefficient (NTC) behavior of decalin autoignition was observed, for the first time, at temperatures below 920 K. Current ignition delay data are in good agreement with past shock tube data in terms of pressure dependence but not equivalence ratio dependence. Detailed comparisons of these ignition delay v

6 data with predictions based on currently available decalin reaction mechanisms are presented, and preliminary suggestions for the adjustment of some key rate parameters are made. Finally, it is found that ignition delay times behind reflected shock waves are strongly sensitive to variations in temperature and pressure, yet most current models of reaction kinetics do not properly account for the variations that are typically observed in shock tube experiments. Particularly at low reaction temperatures with relatively long ignition delay times, substantial increases in pressure and temperature can occur behind the reflected shock even before the main ignition event, and these changes in thermodynamic conditions of the ignition process have proved difficult to interpret and model. To obviate such pressure increases, we applied a new driven-gas loading method that constrains the volume of reactive gases, thereby producing nearconstant-pressure test conditions for reflected shock measurements. Using both conventional operation and this new constrained-reaction-volume (CRV) method, we have collected ignition delay times for 1-butanol/O 2/N 2 mixtures over temperatures between 716 and 1121 K and nominal pressures of 20 and 40 atm for equivalence ratios of 0.5, 1.0, and 2.0. The equivalence ratio dependence of 1-butanol ignition delay time was found to be negative when the oxygen concentration was fixed, but positive when the fuel concentration was held constant. Ignition delay times with strong pre-ignition pressure increases in conventional-filling experiments were found to be significantly shorter than those where these pressure increases were mitigated using the CRV strategy. The near-constant-pressure ignition delay times provide a new database for low-temperature 1-butanol mechanism development independent of non-idealities caused by either shock attenuation or pre-ignition perturbations. Comparisons of these near-constantpressure measurements with predictions using several reaction mechanisms available in the literature were performed. To our knowledge this work is first of its kind that systematically provides accurate near-constant-enthalpy and -pressure target data for chemical kinetic modeling of undiluted fuel/air mixtures at engine relevant conditions. Measurements of stable decomposition products of engine fuel pyrolysis were performed behind reflected shock waves at temperatures from 1100 to 1600 K and pressure of 12 atm for all engine fuels tested. Species time-history measurements of ethylene and methane were made in pyrolysis experiments using laser absorption. Similar product yields were found for conventional fuels but not for alternative fuels at current conditions. These data provide needed input for the development and validation of a new compact hybrid kinetic model under development at Stanford. A traditional jet fuel chemical kinetics model using surrogate fuel vi

7 compounds and combined detailed mechanisms of these surrogates was also discussed and compared with the hybrid model. The performance of the hybrid model appears better than that of the surrogate model, but still needs improvement. Ethylene mole fraction and fuel absorbance time-histories were acquired using laser absorption at 10.6 and 3.39 μm during decalin pyrolysis for mixtures of ppm decalin/argon at pressures of atm and temperatures of K. All current studies of kinetics of conventional engine fuels, engine fuel component (decalin), alternative engine fuels, and alternative fuel component (butanol) provide useful kinetic targets for the development, validation, and refinement of combustion kinetics models, and provide valuable insights on the pre-ignition phenomenon typically observed in shock tube ignition delay experiments. vii

8 viii

9 Acknowledgement I would like to express my gratitude towards my advisor, Professor Ronald K. Hanson, for his support, guidance and motivation over the last six years of my life. He offered me once-in-alifetime opportunity to perform world class research and to study at Stanford University, and taught me to be a careful experimenter and an independent thinker. I would also like to thank Dr. David Davidson for his support and advice throughout my PhD career on the shock tube land. I am also thankful to Profs. Bowman and Wang for participating on my reading committee and my oral examination committee, and Prof. Kovscek for being the chair of my oral examnination committee. Stanford is a remarkable place and the Hanson Lab is an incredible group of students. I am fortunate to spend my time studying and working here. I must convey my deepest gratitude to my mentor in the first few years, Matthew Campbell, who helped me so much from when nothing is working to when experiments are finally going well. I would also like to thank Brian Lam who taught me everything about the high pressure shock tube. Thanks to Megan MacDonald, Wei Ren, Sijie Li, Ian Schultz, Ritobrata Sur, Marcel Nations Martin, Shengkai Wang, Tom Parise, Jiankun Shao, and Yu Wang for many conversations about both lab and life. Thanks also to Rui Xu for plenty of help about the HyChem model used in my work. I could not have asked for a better group of friends with whom to have fun during my time here. I am grateful to my friends, in particular - Yu He, Cong Meng, Yue Ma, Linxiao Zhu, and Kun Yang who have been assisting me in achieving a good work-life balance over the last few years. Finally, and most importantly, I would like to thank my mom, dad, and girlfriend Bing. My family have been a constant anchor and support in my times of trouble, regardless of the geographical distance and time zone difference. I met Bing in 2011 and my PhD studies miraculously turned better after she became my girlfriend. I am blessed to share joy with her. ix

10 x

11 Table of Contents Abstract... v Acknowledgement... ix Table of Contents... xi List of Tables... xv List of Figures... xvii 1 Chapter 1 Introduction Motivation Background Engine Fuels Studied Experimental Conditions Spanned Ignition Studies of Engine Fuels Pyrolysis Studies of Engine Fuels Kinetic Studies of Decalin Ignition Studies of 1-Butanol Organization Chapter 2 Method Shock Tube Laser Diagnostics Non-reactive Shock Mixture Monitoring Chapter 3 Ignition Studies of Engine Fuels Example Measurements Data and Correlations HP Air IDT Data Previous HP Air IDT Data xi

12 3.2.3 LP Air IDT HP 4% O 2/Ar IDT Data HP N 2 vs Ar IDT Data Variable X O2 IDT Data Variable ϕ IDT Data Modeling Surrogate Approach HyChem Approach Comparison of Surrogate and HyChem Simulations Evolution of the HyChem Mechanism Chapter 4 Pyrolysis Studies of Engine Fuels Category-A Jet Fuels Example Measurements Product Yields Evolution of Data Category-C Jet Fuels Example Measurements Product Yields RP-2 Fuels Example Measurements Product Yields Chapter 5 Kinetic Studies of Decalin Ignition Delay Time High-temperature Ignition Low-temperature Ignition Pyrolysis xii

13 μm Diagnostic μm Diagnostic Understanding the Decalin Kinetics Pyrolysis Kinetics Oxidation Kinetics Chapter 6 Ignition Studies of 1-Butanol Constrained-reaction-volume strategy Experimental procedure Mixing characterization Conventional-filling experiments Interpretation of conventional-filling ignition data Constrained-reaction-volume experiments Interpretation of CRV ignition data Chapter 7 Summary and Future Work Engine fuels Decalin Butanol Appendices A1: Tables of Raw IDT Data A2: Tables of Product Yields A3: 3.39 µm Absorption Cross Sections of Fuels A4: Shock Attenuation Issue in HPST CRV Introduction Theory Experiments Conclusions and Future Work xiii

14 8.5 A5: Fuel Specifications Category-A jet fuels Category-C jet fuels RP-2 fuels NASA polynomials thermochemical data for Category-A jet fuels, Category-C jet fuels and RP-2 fuels [58, 59] Bibliography xiv

15 List of Tables Table 1.1 Previous example combustion studies of alternative fuels... 2 Table 1.2 Basic properties of all neat engine fuels studied. More information can be found in Appendix A Table 1.3 Previous experimental studies of pyrolysis of decalin... 9 Table 1.4 Previous experimental studies of oxidation of decalin... 9 Table 1.5 Decalin reaction mechanisms discussed in this study... 9 Table 2.1 Wavelengths and laser types for some chemical kinetics target species Table 2.2 Fuel loss analysis for A2, C1, and C Table 3.1 Summary of ignition delay time data of neat conventional fuels Table 3.2 Summary of ignition delay time data of alternative fuels. Not all fuel blends are included Table 3.3 Datasets of Category-A fuels in Air and Airgon Table 3.4 Some combustion property targets used in surrogate formulation for real fuels. Data were taken from [10, 46, 49, 50] Table 3.5 Screening of combination of surrogate compounds and reaction mechanism for surrogate modeling of J11 IDT (not an exhaustive effort) Table 3.6 Surrogate compositions and reaction mechanisms for fuels in this study Table 3.7 The HyChem model versions [150] Table 4.1 Carbon fractions captured in C 2H 4 and CH 4 diagnostics during pyrolysis of Category- C fuels at near 1600 K and 0.5 ms Table 5.1 Absorption cross sections of major hydrocarbons at 3.39 and μm in the decalin pyrolysis system. Fits listed here were derived from cross section measurements at high T and P Table 8.1 Ignition Delay Times of A Table 8.2 Ignition Delay Times of A Table 8.3 Ignition Delay Times of A Table 8.4 Ignition Delay Times of C Table 8.5 Ignition Delay Times of C Table 8.6 Ignition Delay Times of R Table 8.7 Ignition Delay Times of R Table 8.8 Ignition Delay Times of C xv

16 Table 8.9 Ignition Delay Times of C Table 8.10 Ignition Delay Times of C Table 8.11 Ignition Delay Times of C Table 8.12 Ignition Delay Times of J Table 8.13 Ignition Delay Times of C Table 8.14 Ignition Delay Times of J11/C7 blend (50:50 v.%) Table 8.15 Ignition Delay Times of B Table 8.16 Ignition Delay Times of J11/B12 blend (50:50 v.%) Table 8.17 Ignition Delay Times of S Table 8.18 Ignition Delay Times of S Table 8.19 Ignition Delay Times of J11/S8 blend (50:50 v.%) Table 8.20 Ignition Delay Times of J11/S9 blend (50:50 v.%) Table 8.21 Ignition Delay Times of H Table 8.22 Ignition Delay Times of J11/H10 blend (50:50 v.%) Table 8.23 Ignition Delay Times of H Table 8.24 Ignition Delay Times of J11/H11 blend (50:50 v.%) Table 8.25 Ignition Delay Times of F Table 8.26 Ignition Delay Times of H Table 8.27 Ignition Delay Times of B Table 8.28 Ignition Delay Times of B Table 8.29 Ignition Delay Times of HRA Table 8.30 Ignition Delay Times of K Table 8.31 Ignition Delay Times of K Table 8.32 Ignition Delay Times of K Table 8.33 Ignition Delay Times of decalin in air Table 8.34 Ignition Delay Times of 1-butanol/O 2/N Table 8.35 Product yields during pyrolysis of A1, A2, and A3 (SK3 dataset) Table 8.36 Product yields during pyrolysis of C1, C2, C3, C4, C5, and C Table 8.37 Product yields during pyrolysis of R4 and R Table 8.38 Properties of staged-filling gas (gas 1) and test gas (gas 1 ) at 1 atm and 91 o C. 145 xvi

17 List of Figures Figure 1.1 Experimental coverage of temperature-pressure space of this work of engine fuels. 6 Figure 1.2 Previous ignition delay time measurements 1-butanol oxidation in air at ϕ = 1.0. The constant-energy (U), constant-volume (V) model calculations utilize five recent mechanisms Figure 2.1 Schematic of a shock tube with driver insert and PCB system Figure 2.2 Example absorption cross-section data. Refs: Ren et al. [132], MacDonald et al. [51], Sur et al. [133] Figure 2.3 Schematics of laser-absorption diagnostics set-up used in current ignition studies (LEFT) and pyrolysis studies (RIGHT) Figure 2.4 Example absorbance (and pressure) time-histories in current pyrolysis studies using the set-up shown in Figure 2.3 on the right. Reflected shock condition: 0.6% C6 fuel in argon, 1387 K, 12.3 atm. The inferred species mole fraction time-histories are shown in Figure Figure 2.5 Example non-reactive reflected shock experiment with argon at 1390 K and 17.3 atm. Pressure signal and transmitted signal of CO 2 laser are included Figure 2.6 Schematic of mixing tank and mixing manifold for HPST. The fuel flask was not used in current studies. The gas bottles were prepared based on the experiment needs (i.e., not limited to 21% O 2/Ar and N 2) Figure 2.7 Injection Curve of A2, C1, and C5. The vapor pressure of A2 is 140 Torr at 120 o C Figure 3.1 Ignition delay time measurements using pressure (in black), 306 nm OH* emission (in red), and 3.39 µm He-Ne absorbance (in blue) during A2 oxidation. Initial reflected shock conditions: 1278 K, 15.3 atm, 4%O 2/Ar, equivalence ratio (ϕ) equal to 1.0 for top panel; 1031 K, 11.1 atm, ϕ = 1.0 for bottom panel. The non-reactive pressure profiles at similar conditions are also shown (in green) Figure 3.2 Ignition delay time data for distillate fuel/air, ϕ = , normalized to 12 atm using the pressure dependence of Eqn. 1. The Eqn. 1 correlation is shown as a solid line Figure 3.3 Ignition delay time data for C1/air, ϕ = , normalized to 12 atm using the scalings of Eqn. 2. The Eqn. 2 correlation is shown as a solid line Figure 3.4 Ignition delay times in air of six Category-C fuels at 12 atm and ϕ = 1.0 (solid dots). Dashed line is the correlation Eqn xvii

18 Figure 3.5 Reproducibility of current A1 - A3 HP Air IDT measurements Figure 3.6 Influence of thermochemistry on C1 data report Figure 3.7 Ignition delay time data for previously published fuel/air data, ϕ = , normalized to 12 atm using pressure dependence of Eqn. 1. The Eqn. 1 correlation is shown as a solid line Figure 3.8 Ignition delay time data for J11/air, ϕ = , normalized to 3 atm using the scalings of Eqn. 3. The Eqn. 3 correlation is shown as a solid line Figure 3.9 Comparison of ignition delay time for J11 (JP , shown in dashed line), C7 (Gevo ATJ, shown in red dots), and B12 (Swedish BioJet, shown in blue dots) in air at 3 atm and ϕ = Figure 3.10 Comparison of ignition delay time for J11 (JP ) and six alternative aviation fuels in air at 6 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn Figure 3.11 Comparison of ignition delay time for J11, Swedish BioJet, and 50/50 blend of J11 and Swedish BioJet in air at 3 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn Figure 3.12 Comparison of ignition delay time for J11, SHELL SPK, and 50/50 blend of J11 and SHELL SPK in air at 6 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn Figure 3.13 Comparison of JP-8 and Gevo Air IDT at various pressures Figure 3.14 Ignition delay time data for fuel/4% O 2/Ar data, ϕ = , normalized to 12 atm using pressure scaling of Eqn. 4. The Eqn. 4 correlation is shown as a solid line. Earlier data in 2013 are shown with open symbols Figure 3.15 Comparison of measured IDT of A2 (in green), C1 (in black), and C5 (in red) in two kinds of bath gas, i.e. 4% O 2/Ar (hollow dot with a cross) and air (solid dot), at 12 atm and ϕ = 1.0. Solid lines are best fits to data points Figure 3.16 Comparison of simulated IDT of ethylene (in green) and iso-butene (in black) in two kinds of bath gas, i.e. 4% O 2/Ar (hollow dot with a cross) and air (solid dot), at 12 atm and ϕ = 1.0. Solid lines are best fits. Simulations done using USC Mech II [141] and constant UV assumption Figure 3.17 Comparison of measured IDT for F15 (F-76), H13 (F-76/HRA Blend), B14 (DIM), and B15 (103-2) in 4%O 2/Ar at 17 atm and ϕ = 1.0. Correlation using Eqn. 4 is shown in dashed line Figure 3.18 A2 Air and Airgon IDT with ϕ = only xviii

19 Figure 3.19 Pressure trace in A2 Air and Airgon IDT experiments Figure 3.20 A comparison of A2 jet fuel ignition delay time data for ϕ = 1 for varying amounts of O 2 in Ar and N 2. The 500ppm fuel/o 2/Ar data is J11 POSF 6169, which is very similar to the A2 fuel, and has 0.82% O 2 for ϕ = Figure 3.21 Variation of ignition delay time with equivalence ratio for A1 jet fuel Figure 3.22 Variation of ignition delay time with equivalence ratio for F-76 in 4%O 2/Ar Figure 3.23 J11 IDT in air. Data are shown in solid dots and kinetic modeling (see Table 3.6) in dashed lines Figure 3.24 Comparison of IDT for JP-8, SHELL SPK, and 50/50 blend of JP-8 and SHELL SPK in air at 6 atm and ϕ = 1.0. Data are shown in solid dots and kinetic modeling (see Table 3.6) in dashed lines Figure 3.25 F15 IDT in 4%O 2/Ar. Data are shown in solid squares and kinetic modeling (see Table 3.6) in dashed lines Figure 3.26 RP-1 IDT in 4%O 2/Ar. Data are shown in solid squares and kinetic modeling (see Table 3.6) in dashed lines Figure 3.27 A1 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.28 A2 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.29 A3 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.30 C1 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.31 C5 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.32 R4 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.33 R5 4%O 2/Ar and Air IDT data and HyChem simulations Figure 3.34 A1 IDT in 4%O 2/Ar and Air. Simulations using the HyChem approach are shown in Red and simulations using the surrogate approach (see Table 3.6, surrogate using [49] and mechanism using [53]) are shown in Black Figure 3.35 Comparison of simulated pressure traces in Air and Airgon using the surrogate approach (LEFT) and the HyChem approach (RIGHT) Figure 3.36 Air IDT of A2 using version 1 and version 2 of the HyChem mechanism Figure 4.1 A series of C 2H 4 measurements at different temperatures at 12 atm during pyrolysis of 0.7% A2 in argon Figure 4.2 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during A2 pyrolysis. Smooth solid lines: simulations using the HyChem approach [150] (version 02, May 2016). Initial reflected shock conditions: 1228 K, 12.4 atm, 0.73% A2/argon xix

20 Figure 4.3 Laser absorption measurements of C 2H 4 during J11 pyrolysis. Initial reflected shock conditions: 1231 K, 14.6 atm, 0.26% J11/argon. Smooth dashed lines: simulations using the J11 surrogate model (see Table 3.6) Figure 4.4 C 2H 4 (blue) and CH 4 (red) product yields and associated carbon fractions during Category-A fuel pyrolysis; yields at 0.5 ms in TOP panel, yields at 1.0 ms in MIDDLE panel, and yields at 1.5 ms in BOTTOM panel. Dots represent individual experiment. Solid lines: HyChem simulations. Average reflected shock pressure and fuel loading: 12.4 atm and 0.73% fuel in argon Figure 4.5 Comparisons of online-only and offline-corrected C2H4 yield time history. Initial reflected shock condition: LEFT, 0.75%A2/Ar, 1180 K, 16.3 atm; RIGHT, 0.69% A2/Ar, 1370 K, 15.4 atm Figure 4.6 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during C6 (a), C5 (b), and C1 (c) pyrolysis. Simulations for C5 and C1 using the version 2 of the HyChem model are shown in smooth solid lines. Initial reflected shock conditions: 1387 K, 12.3 atm, 0.6% C6/Ar; 1274 K, 12.8 atm, 0.65% C5/Ar; 1375 K, 14.3 atm, 0.71% C5/Ar Figure 4.7 C 2H 4 (shown in the top panel in squares) and CH 4 (shown in the bottom panel in circles) product yields during Category-C fuel pyrolysis; yields at 500 µs. Average reflected shock pressure and fuel loading: 12 atm and 0.6% fuel in argon. The solid lines are simply smoothed fits to the data Figure 4.8 C 2H 4 and CH 4 product yields during C1 and C5 fuel pyrolysis; yields at 500 µs. Data are shown in squares to C 2H 4 and circles for CH 4, respectively. Modeling was done using the version 2 of the HyChem mechanism with C 2H 4 yield modeling shown in solid lines and CH 4 yield modeling shown in dashed lines. Average reflected shock pressure and fuel loading: 12 atm and 0.6% fuel in argon Figure 4.9 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during R4 pyrolysis. Simulations using the version 2 of the HyChem model are shown in smooth solid lines. Initial reflected shock conditions: 1387 K, 12.8 atm, 0.68% R4/Ar Figure 4.10 C 2H 4 and CH 4 product yields during R4 (shown in black) and R5 (shown in red) pyrolysis. Dots represent individual experiment with solid squares for C 2H 4 and hollow circles for CH 4. Solid lines: HyChem (version 2) simulations for C 2H 4 yields. Dashed lines: HyChem (version 2) simulations for CH 4 yields. Average reflected shock pressure and fuel loading: 13.5 atm and 0.76% fuel in argon xx

21 Figure 5.1 Example ignition delay time measurement at high temperature. The reaction mechanism used for simulating this measurement was published in [82] Figure 5.2 Example ignition delay time measurements at low temperature. The reaction mechanism used for simulating this measurement was published in [82] Figure 5.3 Comparison of ignition delay data of stoichiometric decalin/air mixture at various pressures from RPI [96] and SU (current studies). Solid and dashed lines are best fits to data points Figure 5.4 Comparison of ignition delay data of decalin/air mixture at various equivalence ratios from RPI [96] at 12 atm and SU (current studies) at 20 atm. Solid and dashed lines are best fits to data points Figure 5.5 Comparison of ignition delay data of decalin/air mixture at various equivalence ratios from RPI [96] at 40 atm and SU (current studies) at 50 atm. Solid and dashed lines are best fits to data points Figure 5.6 Ignition delay time data for decalin/air at above 1000 K. All data are normalized to 20 atm and ϕ = 1.0 using the scalings of Eqn. 5. The Eqn. 5 correlation is shown as a solid line Figure 5.7 Comparison of measured and originally predicted ignition delay data of stoichiometric decalin/air mixture at 12, 20, and 50 atm over K. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption Figure 5.8 Comparison of measured and originally predicted ignition delay data of decalin/air mixture at ϕ = 0.5, 1.0, and 2.0, 20atm over K. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption Figure 5.9 Comparison of measured and originally predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 20 atm; ϕ = 1.0, 20 atm; ϕ = 1.0, 50 atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. NTC behavior is evident at around K Figure 5.10 Temperature-dependent pressure and equivalence ratio scaling dimensions for ignition delay times at K obtained from Ranzi modeling Figure 5.11 Example kinetic modeling of decalin pyrolysis at a representative shock tube condition using the reaction mechanism in [82] and constant U, V assumption xxi

22 Figure 5.12 Absorbance time-histories at 3.39 μm during pyrolysis of 2260 ppm decalin/ar at K and atm Figure 5.13 Comparison of measured and originally predicted ethylene yield time-histories during pyrolysis of ppm decalin/ar at K and atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption Figure 5.14 Comparison of measured, predicted and estimated peak ethylene yield during pyrolysis of ppm decalin/ar at K and atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. Upper-limit estimation were derived based on analysis of the reaction pathways in Figure 5.15 (see text) Figure 5.15 Initial decomposition pathways of decalin adapted from [99] (using the same reaction labels). Only pathways 1-1 and 12-1 were included in the estimation of upper-limit of peak ethylene yield Figure 5.16 A-factor sensitivity analysis for ethylene during pyrolysis of 3493 ppm decalin in Ar at 1412 K and 18.8 atm Figure 5.17 Comparison of measured and newly predicted ethylene yield time-histories during pyrolysis of ~3540 ppm decalin/ar at K and atm. Modified Ranzi modeling was made using the reaction mechanism [18] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption Figure 5.18 Comparison of measured and predicted peak ethylene yield during pyrolysis of ~3540 ppm decalin/ar at K and atm. Modified Ranzi modeling was made using the reaction mechanism [18] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption Figure 5.19 Comparison of peak ethylene yields of different classes of alkanes: n-dodecane [76], methylcyclohexane [76], iso-cetane [76], and decalin (this study) Figure 5.20 Comparison of calculated (black) and measured (red) absorbances at μm during pyrolysis of ~3540 ppm decalin/ar at K and atm. Calculations were made using the reaction mechanism [82] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. Figures at right show zoomed-in view of those at left Figure 5.21 Comparison of calculated (black) and measured (red) absorbances at 3.39 μm during pyrolysis of ~2260 ppm decalin/ar at K and atm. Calculations were xxii

23 made using the reaction mechanism [82] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. Figures at right show zoomed-in view of those at left Figure 5.22 Brute-force sensitivity analysis of ignition delay time of decalin in air at 20 atm (TOP) and 50 atm (BOTTOM), ϕ = Figure 5.23 Comparison of currently measured and newly predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 20 atm; ϕ = 1.0, 20 atm; ϕ = 1.0, 50 atm. Modified Ranzi modeling was made using the reaction mechanism [82] with modified rate constant for reaction oxi 4 in Figure 5.22 and constant U, V assumption Figure 5.24 Comparison of previously measured (Oehlschlaeger et al. [96]) and newly predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 12 atm; ϕ = 1.0, 12 atm; ϕ = 1.0, 40 atm. Modified Ranzi modeling was made using the reaction mechanism [82] with modified rate constant for reaction oxi 4 in Figure 5.22 and constant U, V assumption Figure 6.1 Idealized schematic of the gas mixtures during the staged-filling process Figure 6.2 Example monitoring of the pressure in the driven section, the 3.39 μm He-Ne absorbance and the derived 1-butanol concentration in the test gas during the staged-filling process, for staged filling with L 1 = 32 cm. X fuel,manometric, X fuel, 3.39 μm and X fuel, R1 stand for the fuel concentration prepared manometrically in the mixing tank, the one measured in situ in the shock tube at the end of first filling stage, and the one at the end of second filling stage, respectively Figure 6.3 Example monitoring of the 3.39 μm He-Ne laser absorbances and the derived 1- butanol concentrations in Regions 1, 2, and 5 (X fuel, R1, X fuel, R2, and X fuel, R5) after the stagedfilling process for a very lean (ϕ = 0.15) reflected shock wave experiment (shown in Figure 6.2). The slow rise immediately behind the reflected shock is associated with the reflected-shock bifurcation at the sidewall Figure 6.4 Normalized temperature-dependent pressure profiles for ignition delay time measurements of 1-butanol in air at ϕ = 1.0 and near 20 atm with conventional-filling Figure 6.5 Normalized pressure-dependent pressure profiles for ignition delay time measurements of 1-butanol in air at ϕ = 1.0 with conventional-filling Figure 6.6 Comparison of current ignition delay times of 1-butanol in air with previous results at ϕ = 1.0 and 20 atm with conventional-filling xxiii

24 Figure 6.7 Comparison of current ignition delay times of 1-butanol in air with previous results at ϕ = 1.0 and 40 atm with conventional-filling Figure 6.8 Comparison of current conventional-filling ignition delay times of 1-butanol in air at ϕ = 1.0 and 20 atm with modeling results using five recent mechanisms [103, ] Figure 6.9 Normalized pressure traces at five measurement locations for a 1-butanol/air mixture at ϕ = 1.0 with conventional shock tube operation Figure 6.10 Measured, specified, and simulated pressure profiles for an example ignition delay time measurement with conventional-filling. The simulated pressure profile was from calculations using the reaction mechanism of Sarathy et al. [111] and a constant U, V assumption Figure 6.11 Comparison of current conventional-filling ignition delay data with two model predictions using both constant U, V assumption and the specified-pressure method Figure 6.12 Comparison of conventional-filling and CRV ignition delay data at 1047 K, -20 atm, ϕ = Figure 6.13 Comparison of conventional-filling(left) and CRV(RIGHT) ignition delay data at -880 K, -20 atm, ϕ = Figure 6.14 Example CRV measurements of ignition delay time at low temperatures for 3.38% 1-butanol/20.3% O 2/N 2 (LEFT) and 3.38% 1-butanol/40.6% O 2/N 2 (RIGHT) Figure 6.15 Comparison of conventional-filling and CRV ignition delay times of 1-butanol in air at ϕ = 1.0 and 20 atm. Solid lines are simply best fits to data Figure 6.16 Comparison of current CRV data at ϕ= 1.0, 20 atm with model predictions at constant H, P Figure 6.17 Comparison of current CRV data and Sarathy et al. [111] mechanism predictions at ϕ = 0.5, 1.0, 2.0, 20 atm with oxygen concentration being fixed (-20%) Figure 6.18 Comparison of current CRV data and Sarathy et al. [111] mechanism predictions at ϕ = 0.5, 1.0, 2.0, 20 atm with fuel concentration being fixed (3.38%) Figure 8.1 Temperature-dependent absorption cross section of B12, C7, J11, and J11/C7 Blend at 3.39 μm Figure 8.2 Temperature-dependent absorption cross section of B14, F15, H13, and B15 at 3.39 μm Figure 8.3 Temperature-dependent absorption cross section of A1, A2, and A3 at 3.39 μm. 141 Figure 8.4 Temperature-dependent absorption cross section of decalin at 3.39 μm. Representative error bars are shown xxiv

25 Figure 8.5 Comparison of velocity measurements from two different staged-filling locations, i.e., 200 cm and 496 cm from driven section endwall, respectively Figure 8.6 Left-hand-side (LHS) and right-hand-side (RHS) of Eqn. (9) Figure 8.7 Comparison of velocity measurements for various concentrations of CO 2 in the staged-filling gas mixture of CO 2 and N Figure 8.8 Variation of overall attenuation rate as a function of initial concentration of CO 2 in the staged-filling gas Figure 8.9 Comparison of ignition delay times from new and old CRV experiments Figure 8.10 Comparison of all CRV data and kinetic predictions using three reaction mechanisms xxv

26

27 1 Chapter 1 Introduction 1.1 Motivation With the decline of supplies of conventional petroleum-based fuels, alternative and renewable fuel sources will be needed to provide energy and power for future generations. Emerging fuels include bio-derived fuels that hold the potential to become carbon neutral and help mitigate the greenhouse gas emissions implicated in global climate change [1, 2]. Meanwhile, engine technologies will advance to make more fuel-efficient and environment-friendly engines that would potentially consume non-petroleum-derived fuels. As a result, in recent years, for both strategic and environmental reasons, alternative fuel strategies to replace or supplement petroleum-derived practical fuels (such as JP-8 for aviation, and F-76 for navy applications) have been proposed to meet military and civilian energy goals [3, 4]. For example, in the U.S., the Federal Aviation Administration (FAA) is working to enable the use of one billion gallons per year of drop-in sustainable alternative jet fuels by 2018 [5]. In the world, a total of 21 airlines have now used alternative fuel for commercial flights (as of December 2014, [5]). The U.S. Navy set the goal in 2012 of 50% use of alternative energy by 2020 for their great green fleet [6]. It is generally believed that the use of alternative (non-petroleum-derived) fuels can help address the challenges imposed by conventional (petroleum-derived) fuels, such as fuel cost, environmental impact, energy security, etc.. And it is generally found that each type of alternative fuel has its own unique manufacturing feedstock, physical characteristics and chemical signatures; biomass-derived fuels from celluloses and lignins have high polymeric characteristics, oil sands fuels have unusually high naphthenic content, and oil shale and coal liquid fuels have more aromatic content than conventional petroleum-based engine fuels, just to name a few [1]. This historic opportunity of transitioning to alternative fuels and advanced engines urges fundamental research in gas phase chemistry, multiphase chemistry and dynamics, and multiscale modeling and simulation [1, 2]. These studies have extended to university laboratories; some examples are listed in Table

28 Table 1.1 Previous example combustion studies of alternative fuels Fuel Type Reference Coal-derived iso-paraffinic kerosene (IPK) [7, 8] Fischer-Tropsch (FT) synthetic paraffinic kerosene (SPK) [7] Hydro-processed renewable jet (HRJ) derived from camelina and tallow [7, 9] Alcohol-to-jet fuels (ATJ) [10] Hydro-refined algal oil [6] Terpene dimers [11] Hydrogenated-pinene [11] Gas-to-Liquid (GtL) FT synthetic kerosene [12, 13] Coal-to-Liquid (CtL) [14] Farnesane (trimethyl dodecane) [15] Category-C jet fuels C-2 - C-5 [16] Syntroleum S-8 [8, 17] R-8 [17] Related to the studies of these multi-component engine fuels themselves, several individual single-component fuels have drawn much attention recently. For example, decalin is a bicyclic alkane composed of two fused six-membered rings and occurs in both cis and trans isomers. It is a primary constituent of petroleum feedstocks and found in automotive fuels and proposed additive packages for these various fuels [18]. It has also been found to be significant in the new generation of fuels derived from tar sands and oil shales [19-22]. Additionally, it is claimed that decalin has effective endothermic fuel capability particularly attractive to engine cooling [18, 23, 24], and thus a good knowledge of vapor-phase reaction kinetics of decalin pyrolysis and combustion is needed [25]. Decalin has also been chosen as an archetypical cyclo-alkane class component for surrogate jet fuels [26] and surrogate diesel fuels [27, 28]. In spite of this, Granata et al. [29] stated that cyclo-alkanes, including decalin, have received scant attention, and kinetic knowledge of their combustion is less defined and accurate. All of these characteristics warrant the experimental and theoretical investigations of decalin combustion reaction kinetics. Similarly, there has been strong interest in the combustion of 1-butanol and other butanol isomers due to their potential as biofuel candidates for next-generation green transportation fuels [30, 31]. Butanol ignition data at high pressures and low temperatures are particularly needed due to their relevance to practical engine combustion environments. All of this motivates 2

29 current work of conventional engine fuels, alternative engine fuels, decalin (conventional engine fuel component), and 1-butanol (alternative engine fuel component). In support of the development and deployment of sustainable alternative engine fuels, many university laboratories have been working on the development of accurate and reliable chemical kinetics models for these fuels. Traditionally, a surrogate approach has been employed to facilitate the development of these models by selecting representative hydrocarbons to mimic the real fuels [9, 16, 26-28, 32-55]. However, since many of the proposed new alternative fuels contain species with large molecular weight and with diverse compounds such as oxygenates, naphthenes and other types of species, whose physical and chemical properties and combustion kinetics are poorly known, the surrogates and/or detailed reaction mechanisms of most alternative fuels are still in their infancy or even do not exist. To exacerbate the situation, with the ever-expanding alternative jet fuels industry, this surrogate approach appears slow and inefficient, and inevitably leads to the convolution of coupling of combustion chemistry mechanism development with mechanisms for individual surrogate fuel components. Therefore, to produce a more compact chemical kinetics model for jet fuel that accurately reproduces the pyrolysis and oxidation behavior of real jet fuel, more recently, a new approach, named HyChem, has been suggested by Wang [56]. In this approach, an experimentally-constrained pyrolysis detailed oxidation model can provide both a small reaction mechanism and a direct link to real fuels. In either fashion, the development of kinetic models to describe the combustion behavior of conventional and alternative fuels requires a set of kinetic targets (such as ignition delay times and species time-histories) to test, refine and validate the reaction mechanism. For conventional engine fuels, fundamental experimental databases that might be used for comparison and validation purposes for the models are scarce. There is an even greater lack of data for alternative fuels [1]. In the specific realm of gas phase chemistry, too little is known about the combustion kinetics of conventional or alternative fuels. Shock tube/laser absorption methods can furnish reliable, economical and accurate kinetic targets/tests of ignition, pyrolysis and oxidation that are needed to both validate the reaction mechanisms and characterize the chemical and physical fit-for-purpose properties for these fuels. Experimentally, evaluation of the fuel decomposition product distribution can be measured in flow reactor studies at long times (e.g. 5+ ms). However, complications in flow reactor studies because of the initial fuel mixing time, and in some cases vitiated flow conditions, preclude 3

30 drawing conclusions from the flow reactor measurements at shorter times. Species concentration measurements relevant to actual jet engine resident times require time-history data over significantly shorter test times. It is, in the first 500 µs to 1 ms of high temperature engine environment, where the relevant engine chemistry matters; and it is here that measurements, behind reflected shock waves, of species concentration time-histories using laser absorption methods can provide the necessary data. Other laboratory modeling targets include, for example, ignition delay time (IDT), which places global constraints on model predictions, and species time-histories that place more specific constraints on the internal sub-mechanisms of these models. Last but not least, when it comes to screening of alternative fuel candidates, shock tube methods appear to be more cost-effective than Original-Equipment-Manufacturer (OEM) combustor tests. Therefore, this work is concerned with the role of shock-tube/laserabsorption measurements in providing accurate kinetics targets for the development and refinement of gas-phase chemical kinetics models of conventional engine fuels, alternative engine fuels, decalin and 1-butanol. The research background of these fuels will be discussed in later sections. 1.2 Background Engine Fuels Studied This work investigated a large series of neat conventional fuels, neat alternative fuels, and blends of conventional and alternative fuels. Based on their applications, these fuels can be classified into four groups: jet fuel, navy fuel, rocket-propellant (RP) fuel, and kerosene. Table 1.2 lists the name, identifier used in this text, POSF label, molecular weight, nominal chemical formula and some general specifications available for the neat conventional and alternative fuels tested. For all fuels tested, C NH M values, needed for accurately determining gas-phase equivalence ratios in the mixtures, were determined from independent H/C ratio and GCxGC composition measurements [57-59]. More information about these fuels will be found in tables in later sections. 4

31 Table 1.2 Basic properties of all neat engine fuels studied. More information can be found in Appendix A5. Group Name ID Used in Text POSF Label Molecular Weight (MW, g/mol) Nominal Formula General Specifications a, b [60] Jet Fuel JP-8 J C10.9H22.0 n- > iso- > cyclo- > aro SASOL IPK S C10.5H23.0 iso- > cyclo- > n- SHELL SPK S C9.6H21.4 iso- > n- > cyclo- HRJ-Tallow H C11.3H24.6 iso- > n- > cyclo- HRJ-Camelina H C11.6H25.5 iso- > n- > cyclo- GEVO ATJ C C12.2H26.4 N/A Swedish BioJet B C11.0H22.0 N/A Category A-1 A C10.8H21.6 JP-8 (best case, low aromatic) A-2 A C11.4H21.7 Jet-A (average case) A-3 A C12.0H22.3 JP-5 (worse case, low H-content) Category C-1 C C12.5H27.1 Gevo ATJ (100%), extremely low cetane, C12 and C16 iso-paraffins C-2 C C12.3H % C14 iso-paraffins/16% TMB, extremely asymmetric boiling range C-3 C C12.8H % A-3 fuel/36% farnesane, very high viscosity jet fuel C-4 C C11.4H % Sasol IPK/40 % C-1, low cetane fuel, but broader boiling range C-5 C C9.7H % C10 iso-paraffins/27% TMB, extremely flat boiling range C-6 C C11.9H % cycloparaffins, Virent HDO SK Navy Fuel F-76 F-76 Blend F15 H13 N/A N/A C14.8H26.9 C14.6H29.5 N/A 50% F-76/50% Hydro-refined Algal Oil (HRA) DIM B14 N/A 182 C13.4H % Terpene dimers (TD) /50% JP B15 N/A 185 C13.6H % TD/50% Hydrogenated alpha pinene RP Fuel RP-2-1 R C12.0H24.1 YA2921HW10 RP-2-2 R C12.6H25.6 WC0721HW01 RP-1 R6 N/A C12.2H23.4 SH2421LS05 Kerosene K-1 K6 N/A C11.5H25 Kerosene K-2 K7 N/A C12 H23.0 Kerosene K-3 K8 N/A C13.7 H23.0 N0.015 Diesel a n-: n-alkanes, iso-: iso-alkanes, cyclo-:, cyclo-alkanes, aro: aromatics, TMB: trimethyl benzene, TMD: trimethyl dodecane, HRA: hydro-refined algal oil, TD: terpene dimers b blend percentages are volume based 5

32 1.2.2 Experimental Conditions Spanned A wide range of temperature (T) and pressure (P) were targeted in this work on engine fuels. Putting this in perspective, Figure 1.1 shows the T-P parameter space covered for each of the four groups of engine fuels listed in Table 1.2. As shown, most of current measurements, such as for jet fuel and RP fuel, were conducted at above 1000 K. This is in line with the normal operating temperature of the engines that power current vehicles, especially present aeroengines. Additionally, all tests of four groups of engine fuels were mainly focused on elevated pressure conditions (above 10 atm). This is, again, of practical relevance to real engine combustion performance. For simplicity, in later sections, the two rule-of-thumb numbers, i.e K and 10 atm, will be used to split the experimental coverage of T-P space into high-temperature (HT) regime vs. lowtemperature (LT) regime, and high-pressure (HP) regime vs. low-pressure (LP) regime, respectively Temperature (K) Jet Navy RP Kerosene Pressure (atm) Figure 1.1 Experimental coverage of temperature-pressure space of this work of engine fuels Ignition Studies of Engine Fuels A wide literature on ignition delay time measurements in shock tubes exists. For a review of methods and data, for example, see Lifshitz (2001) [61] and Davidson and Hanson (2016) [62]. Previous (before 2010) work on ignition of engine fuels can be found in [44, 63-65]. Recent works on ignition of alternative aviation fuels are reported in [7-9, 17, 66]. 6

33 One concern of model developers is that because distillate fuels have variable compositions, kinetics models and fuel surrogates may have to be tuned to the specific composition of the individual fuel to accurately reproduce critical combustion parameter such as ignition delay times and flame speeds. Based on our current knowledge, for example of the large variation of ignition delay times for archetypal individual surrogate components, this concern is justified [67]. Distillate fuels, however, are not composed of a small and limited number of components as are surrogate fuels. Surrogate fuels are often designed to include a small number of selected individual archetypal components of n-alkane, iso-alkane, cyclo-alkane and aromatic species. GCxGC analyses of distillate fuels, on the other hand, demonstrate an immense array of different component species. While this large variability in composition might be viewed with concern, recent measurements of distillate fuel pyrolysis products have demonstrated that only a small number of common intermediate species form [68], providing the basis for a possible argument that some combustion characteristics of distillate fuels should actually be similar. Related to this, recent work on model development for practical distillate fuels (e.g. the hybrid model of H. Wang et al. [56, 69]) hypothesizes that ignition may be viewed, simplistically, as initial pyrolysis followed by oxidation. With this model, the observed commonality in decomposition products may work in the modeler s favor to simplify the task of generating a common mechanism for all jet fuels. In short, if most fuels generate approximately the same decomposition products, should not the ignition delay times be similar? This question has been difficult to study because there have been relatively few hightemperature ignition delay time studies of distillate fuels. Gauthier et al. measured ignition delay times in gasoline in 2004 [70]. Early jet fuel studies were performed by Dean et al. [71]. Vasu et al. in our laboratory then investigated IDT for a series of jet fuels [44]; this work was continued by Zhu et al. for F-76 [64], and Haylett et al. for DF-2 [65]. Wang and Oehlschlaeger [8] and Dooley et al. [46] provided high-pressure ignition delay time for a specific Jet A fuel (POSF 4658). Liang et al. measured ignition delay times for China #3 aviation kerosene [72]; Steil et al. measured highly dilute kerosene in oxygen/argon [73]. Other workers have studied jet fuel ignition delay times, but without a major emphasis on gas-phase high temperature IDT (see ref. [44]). Recently, Gowdagiri et al. found in a study of F-76 and bio-fuels that highmolecular weight, mostly aliphatic fuels had very similar high-temperature ignition properties [74]. In their study, they discovered an IDT correlation for these fuels at temperatures above 1000 K that captures the weak dependence on equivalence ratio (ϕ) and an approximate 7

34 expected dependence on pressure in atm (P -1 ) and temperature in K (Arrhenius form) for these distillate fuels Pyrolysis Studies of Engine Fuels Few pyrolysis studies of engine fuels exist in the literature. Owing to the coking concern in rocket engines, pyrolysis studies of RP fuels have drawn more attention. For example, MacDonald et al. studied decomposition of two RP fuels (RP-1 and RP-2) and their kinetic surrogates [52, 75, 76]. For jet fuels, less attention is paid to their decomposition kinetics. Nonetheless, most conventional jet fuels can be considered to have three major types of hydrocarbon class components: alkanes, cycloalkanes and aromatics. Current kinetics modeling of kerosenic fuel decomposition indicates that during high-temperature gas-phase pyrolysis, the majority of the carbon in these original large jet fuel molecules quickly converts to a limited series of smaller hydrocarbons. This series includes ethylene C 2H 4, propene C 3H 6, different isomers of butene though primarily isobutene ic 4H 8, methane CH 4, hydrogen and small aromatics such as benzene or toluene. Many of these products can now be measured using shock-tube/laser-absorption methods [77]. Based on the modeling of large alkanes [49] the product species found for different jet fuels, while showing some variation within the official jet fuel specification, are expected to be generally similar. The aforementioned Wang s HyChem model developed using these product species is expected to satisfactorily characterize the fuel pyrolysis process and the subsequent oxidation process Kinetic Studies of Decalin Previous experimental studies of pyrolysis and oxidation of decalin are summarized in Table 1.3 and 1.4, respectively. While these studies appear to have covered a wide range of temperature (T) and pressure (P), few studies exist for pyrolysis of decalin at high temperatures and pressures (e.g., T > 1100 K and P > 17 atm) and for oxidation of decalin at low temperatures and high pressures (e.g., T < 1000 K and P > 17 atm). Low-temperature and high-pressure oxidation of decalin is of particular interest as it holds the potential of discovery of the negativetemperature-coefficient (NTC) behavior [8, 44, 78-81]. To the best of our knowledge, the NTC behavior of decalin oxidation has not been reported yet. In contrast to experimental studies, surprisingly, theoretical investigations of decalin reaction kinetics are scarce. Only three relevant reaction mechanisms are available in literature and summarized in Table 1.5, with the one released in [82] being the only complete kinetic model that can be readily used. 8

35 Table 1.3 Previous experimental studies of pyrolysis of decalin Refs. Cis- or trans-decalin? T (K) P (atm) Type of experiment? Time-resolved? [83] cis (partial) heated quartz reactor No [84] mixture (47:53 wt%) flow reactor No [85] mixture 1053, flow tubular reactors No [86] trans stainless steel flow reactor No [87] mixture (14:86 wt%) , 2 three flow reactors No [88] mixture (as additive) tubular reactor No [89] mixture (40:60 v.%) conventional flow reactor and No cw CO 2 laser [90] trans very low pressure pyrolysismass No spectrometry [91] mixture (37.6:62.4 v.%) flow reactor Yes [92] mixture a Pyrex glass tube reactor and a Yes stainless steel tubing bomb reactor [93] mixture flow reactor No [94] mixture a micro-pyrolysis laboratory reactor No Table 1.4 Previous experimental studies of oxidation of decalin Refs Cis- or trans-decalin? T (K) P (atm) ϕ Type of experiment Type of data [95] mixture , 0.2 reflected shock wave ignition delay time [96] mixture (35:65 wt%) , 1.0 reflected shock wave ignition delay time [97] cis-, trans-, mixture (35:65 wt%) ignition quality tester ignition delay, DCN [36] mixture (35:65 wt%) flow reactor CO production [91] mixture (37.6:62.4 v.%) flow reactor product speciation [43] mixture flow reactor CO production [98] mixture (38:62 wt%) motored engine product speciation Table 1.5 Decalin reaction mechanisms discussed in this study Mechanism Year Type No. of species No. of reactions Zeppieri et al. [25, 91] 1997 pyrolysis sub-mechanism Chae et al. [99] 2007 pyrolysis sub-mechanism Dagaut et al. [82] 2013 complete mechanism Ignition Studies of 1-Butanol Many experimental studies of the oxidation of one of the butanol isomers, 1-butanol, have been performed by researchers using various facilities and techniques [ ]. However, there still 9

36 appears to be a lack of consensus among the ignition delay time measurements found in these studies (see Figure 1.2), especially at low temperatures. In addition, while the pressure dependence of ignition delay time of 1-butanol/air mixture has been studied extensively to yield a consistent result [ ], few investigations exist for the equivalence ratio or oxygen concentration dependence except that reported in [104]. 1250K 1000K 833K 714K t ign (ms) % 1-butanol/air 20 atm (scaled), =1.0 Experiments Weber (RCM, 2011) Heufer (ST, 2011) Stranic (ST, 2012) Modeling (constant UV) 1-Grana et al. (2010) 2-Black et. al. (2010) 3-Merchant et al. (2012) 4-Sarathy et al. (2012) 5-Vranckx et al. (2011) /T (1/K) Figure 1.2 Previous ignition delay time measurements 1-butanol oxidation in air at ϕ = 1.0. The constantenergy (U), constant-volume (V) model calculations utilize five recent mechanisms. Computational studies of 1-butanol ignition are also abundant. Although detailed chemical kinetic reaction mechanisms for 1-butanol and the other isomers have proliferated rapidly in recent years [103, ], Figure 1.2 shows that there are significant discrepancies between the different mechanism predictions, particularly at relatively low temperatures (T < 1000 K). While it is true that some of these mechanisms were not optimized for low temperature chemistry, these discrepancies still imply that more high-quality data at low temperatures are of critical importance for improvement of these mechanisms. This wide variation in model prediction thus motivates further measurements of high-pressure and low-temperature 1-butanol ignition delay times. In some of the previous 1-butanol oxidation studies behind reflected shocks, the mixtures utilized were highly diluted in argon and thus the energy release during reaction was small [100, 107, 108]. However, in many other cases, undiluted high-concentration mixtures (fuel/air) at elevated pressures and low temperatures were employed and these data were found to exhibit pre-ignition pressure ramps and/or steps prior to the primary ignition event [78, 80, 107, ]. The reasons for these rises are difficult 10

37 to confirm as both real fuel chemistry inside the measurement volume and non-localized ignition outside of the reaction volume can contribute to the pressure increases observed inside the measurement volume [107]. Similar phenomena were also observed in rapid compression machines (RCM) [ ] and research engines [124]. Such dramatic pressure effects are different from the normal, though undesired, gradual pressure increase (dp 5/dt) observed behind reflected shocks in dilute mixtures [125], and these effects may be temperature-dependent [103, 126], pressure-dependent [103, 126], and even fuel-dependent [127]. Generally these effects cannot be simply reproduced by a detailed kinetic mechanism under constant-energy (U), constant-volume (V) constraints. These complications are further motivation to seek strategies for interpreting data exhibiting pre-ignition pressure change, as well as to establish new techniques for generating databases without these thermodynamic complications. 1.3 Organization In Chapter 2, the basics of shock tube and laser absorption diagnostics are discussed, followed by a detailed description of mixture monitoring methods aimed at current experimental studies of practical fuels. In Chapter 3, the observation of Gowdagiri et al. [74] about high-molecular-weight fuels is further investigated and summarized utilizing measurements of the ignition delay times for a wide variety of distillate fuels over a wide range of temperatures, pressures, and mixtures. It is found that under certain constraints, these ignition delay times can all be well-correlated with simple relationships. Here, we report measurements of reflected-shock ignition delay times of all the neat fuels listed in Table 1.2 and 50/50 blends of two neat fuels to characterize and compare these fuels for use as targets for kinetics model development. Kinetic modeling for some fuels utilizing appropriate surrogate compositions and reaction mechanisms as well as the HyChem modeling approach mentioned above is also illustrated and the implications of comparison of the relevant data and modeling are discussed. In Chapter 4, the stagnated flow behind reflected shock waves is used to provide the nearconstant high-temperature, high-pressure, test gas conditions for pyrolysis experiments of Category-A, Category-C, and RP fuels, with ethylene and methane species time-history measurements and comparisons of experimental data of three categories of fuels addressed and discussed. Simulations using the version 2.0 of Wang s HyChem model are also included. 11

38 In Chapter 5, the goal is to extend the kinetic database of decalin at elevated pressures including low-temperature-oxidation and high-temperature-pyrolysis using reflected shock wave experiments. The ignition delay time of decalin in air was first measured over a wide range of temperature, pressure, and equivalence ratio to explore potential NTC behavior. Then, the decalin pyrolysis was reported using time-resolved laser-absorption data at 3.39 μm and 10.6 μm to gain kinetic insight regarding decalin decomposition and ethylene formation. Both ignition delay time and species time-histories for decalin provide needed kinetic targets for the validation and refinement of decalin reaction mechanisms. In Chapter 6, the ignition delay times of 1-butanol in mixtures of oxygen and nitrogen were measured behind reflected shock waves at various temperatures, pressures and equivalence ratios, using two methods. First, data were collected using a conventional shock-tube test-gasloading method. These data were simulated with both a constant-volume model and a specifiedpressure model incorporating the measured pressure profiles into the chemical kinetic model calculations. The measurements were then repeated employing a new shock-tube test-gasloading method, called constrained reaction volume (CRV) [128], that limits the volume of reacting gases and creates near-constant-pressure test conditions. These CRV measurements can be compared to mechanism predictions with a simple and appropriate gasdynamic model using constant or specified pressure (P) and enthalpy (H), thereby avoiding the complications and errors associated with constant U, V or constant P, H modeling of conventional (full shock tube filling) reflected shock experiments. 12

39 2 Chapter 2 Method 2.1 Shock Tube A schematic of the shock tube is shown in Figure 2.1. A shock tube is comprised of two sections, a driver and a driven section, separated by a diaphragm. The driven section is filled to the desired pressure with a mixture of fuel and bath gas (Region 1), and the driver is filled with a light gas, often helium, until the diaphragm bursts causing a shock wave to propagate down the tube into the fuel mixture, heating and pressurizing this mixture. The shock then reflects from the endwall of the shock tube and travels back toward the driver section, again increasing the temperature and pressure of the fuel mixture, now to the desired test conditions (Region 5). Diagnostics are located at or near the endwall for observation of this high-temperature, highpressure test gas. Figure 2.1 Schematic of a shock tube with driver insert and PCB system. All experiments were carried out behind reflected shock waves in the Stanford high-pressure shock tube (HPST). The driver section is 3 m long with a 7.5-cm internal diameter. The stainless steel driven section has a length of 5 m and an internal diameter of 5 cm and was heated to 110 o C to prevent condensation of the test gas mixture. The driver and driven sections are separated by scribed aluminum diaphragms whose thickness can be customized. Non-reactive pressure profiles in these experiments (using N 2 or Ar instead of O 2 in the mixture, see Figure 2.5) were adjusted using driver inserts (shown schematically in Figure 2.1) to limit pressure (and hence also temperature) variations (dp 5/dt) to less than 1%/ms over the required test times [129]. The shock tube was normally helium-driven, though a tailored gas mixture (~40%N 2/He) was used as the driver gas for experiments to extend test times up to 10 ms. Five axially-spaced PCB pressure transducers (PCB 113A, see Figure 2.1), at locations 1.1 cm, 31.6 cm, 69.7 cm, 92.5 cm and cm away from the endwall of the driven section, were employed to record pressure during the ignition experiments and connected to four Philips time-interval counters (model PM6666) for measurement of the incident shock speed. The initial reflected shock 13

40 temperature and pressure were calculated from the incident shock speed extrapolated at the endwall, using one-dimensional shock-jump relations and assuming vibrational equilibrium and frozen chemistry, with uncertainties in initial post-shock temperature and pressure of less than ±1.5%. Pressure time-histories in the test section were monitored using a Kistler TM piezoelectric pressure transducer located 1.1 cm away from the endwall. The driven section was evacuated using a turbo-molecular vacuum system to an ultimate pressure of less than 10-5 Torr, with a combined leak and outgassing rate of less than 10-4 Torr/min before each shock. Further details on the shock tube facility can be found in [130]. 2.2 Laser Diagnostics Three diagnostics were employed during the IDT experiments: laser absorption at 3.39 μm, excited OH radical (OH*) emission near 306 nm, and sidewall pressure mentioned above. Initial fuel concentration was measured in situ in the shock tube via laser absorption with a 3.39 μm Spectra-Physics or Jodon Helium-Neon (He-Ne) laser with InfraRed Associates liquid-nitrogencooled InSb detectors (1% absorption sensitivity). The absorption cross section of the engine fuels was measured in the shock tube in separate experiments using neat fuel with known quantities, and these values agreed within 2.5% with cross-section measurements from another apparatus (Fourier Transform Infrared Spectrometer) in the laboratory [131]. Typical uncertainties in the fuel loading based on these absorption cross section have a Standard Error of 1σ = 2-5%. Ignition was indicated by emission near 306 nm from the A 2 Σ+ - X 2 Π ((0,0) band) of OH* that was detected using a modified Thorlabs TM PDA36A Si detector and Schott UG5 filter with simple collection optics that provided a temporal resolution of less than 10 μs. The measurement location of all diagnostics was also 1.1 cm away from the endwall. In pyrolysis experiments, narrow-linewidth laser absorption took advantage of the Beer- Lambert law, i.e. -ln((i/i o) λ) = σ λnl, to relate the measured absorbance -ln((i/i o) λ), to the unknown species concentrations X NRT/P, using measured absorption cross sections σ λ. When one product dominated the absorbance at a particular wavelength, and other species have broad, nearly constant, and featureless absorbance at this wavelength, a two-wavelength differential method, i.e. on-line minus off-line absorbance, was used to determine the concentration of the dominant absorber. This is the case for ethylene, C 2H 4 [132] and methane, CH 4 [133] in this work. The wavelengths and laser types for the species indicated are shown in Table 2.1. For each wavelength, absorption cross-section data was collected for each species. Some example cross-section data are shown in Figure 2.2. Absorption cross-section at 3.39 µm 14

41 of some engine fuels and fuel components can be found in Appendix A3. Wide variation in the absorption cross-sections at each wavelength and for each species occurs; the particular species wavelengths were chosen to maximize the signal for that particular species relative to other absorbers. Table 2.1 Wavelengths and laser types for some chemical kinetics target species Wavelength (µm) Laser Type Usage Interband Cascade Laser Methane on-line Interband Cascade Laser Methane off-line He-Ne Fuel in Region CO 2 Ethylene on-line CO 2 Ethylene off-line Absorption Cross-Section (m 2 /mol) P ~ 13 atm C 2 H m, Ren et al. C 2 H m, MacDonald et al. C 2 H 3.39 m, MacDonald et al. C 2 H m, this work CH m, Sur et al. CH m, Sur et al. CH 3.39 m, this work Temperature (K) Figure 2.2 Example absorption cross-section data. Refs: Ren et al. [132], MacDonald et al. [51], Sur et al. [133]. Simple schematics of laser absorption diagnostics set-ups used in current work on the highpressure shock tube are shown in Figure 2.3, with diagnostics for IDT shown on the LEFT and for pyrolysis on the RIGHT. Window materials were selected to suit the particular laser wavelength and pressure conditions. For example, while Sapphire is sufficiently strong to permit 3.39 µm, µm, µm, and 306 nm diagnostics, it fails to transmit enough light 15

42 longer than 6 µm, so a 3-mm-thick Zinc-Selenide (ZnSe) window with anti-reflection (AR) coating was used in the µm and µm diagnostics at 10 atm. Thermoelectric-cooled IR photovoltaic detectors from Vigo Systems were used for sensing of C 2H 4 and CH 4. It is worth mentioning that because of the high power (300 mw) of the CO 2 laser used for C 2H 4 sensing, the integrating sphere method was feasible in the C 2H 4 diagnostic [134]. Figure 2.3 Schematics of laser-absorption diagnostics set-up used in current ignition studies (LEFT) and pyrolysis studies (RIGHT). Example absorbance time-histories using the set-up for pyrolysis studies are shown in Figure 2.4. Species (C 2H 4 and CH 4 here) mole fractions can be inferred from these absorbance datasets and the cross-section data shown above in Figure 2.2. It should be noted that since cross-section data are usually temperature-dependent (see Figure 2.2), simulated temperature time-histories using available jet fuel reaction mechanisms and constant-pressure (see pressure signal in Figure 2.4), constant-enthalpy assumptions were employed to assist the inference of species mole fractions over time. Besides measuring initial fuel concentration, the 3.39 µm He-Ne laser diagnostic plays an important role in understanding the pyrolysis kinetics. At long times (1 ms and later), it is noted that the 3.39 µm absorbance data appears to reach a plateau, possibly indicating the complete removal of the parent fuel [51]. With cross-section data at 3.39 µm, this long-time relativelyconstant 3.39 µm absorbance data holds the potential of measuring more species, via subtracting off the contributions from the directly measured species (C 2H 4 and CH 4 here) and recognizing the conservation of carbon element and hydrogen element. Due to the variety of pyrolysis fragments, this indirect way of species sensing requires more comprehensive measurements of absorption cross section (i.e. more than just ethylene and methane, such as ethane, iso-butene, etc.), which is ongoing work in our laboratory. Eventually, with more species information, 16

43 distributions of pyrolysis products can be obtained and compared to kinetics modeling, at least at some long times (e.g. 1 ms). Pressure signal or absorbance m m m m m pressure 0.6%C6/Ar 1387 K 12.3 atm Time ( s) Figure 2.4 Example absorbance (and pressure) time-histories in current pyrolysis studies using the setup shown in Figure 2.3 on the right. Reflected shock condition: 0.6% C6 fuel in argon, 1387 K, 12.3 atm. The inferred species mole fraction time-histories are shown in Figure Non-reactive Shock Non-reactive reflected shock wave experiments serve as a good controlled way of checking the quality of the experimental set-up. Figure 2.5 shows an example non-reactive shock wave experiment with pure argon as the driven gas to check the quality of the shock tube facility and the CO 2 laser alignment. First, the pressure signal indicates two things. One is that the test time is approximately 2 ms, and the other is that the driver-insert configuration has been optimized for limiting the dp 5/dt to values effectively near zero within the test time, thereby confirming the constant pressure (and temperature) condition in the absence of reactions. Second, comparing the transmitted signal of CO 2 laser before and during the shock shows that signal remains nearly intact behind the reflected shock, implying that the beam steering effect behind the reflected shock has been mitigated. Similarly, this confirms the constant transmitted signal (or, in the case of common-mode-rejection for unstable lasers, the constant ratio of transmitted signal to incident signal) in the absence of reactions. Such non-reactive shocks were always run 17

44 before real engine fuel tests to make sure that the whole set-up was functional and optimized, laying a solid foundation for collection of high-quality data. 1.5 CO 2 laser signal before shock Signal CO 2 laser signal during shock Pressure test time: ~2 ms pure argon 1390 K 17.3 atm integrating sphere method Time ( s) Figure 2.5 Example non-reactive reflected shock experiment with argon at 1390 K and 17.3 atm. Pressure signal and transmitted signal of CO 2 laser are included. 2.4 Mixture Monitoring Figure 2.6 shows a schematic of the mixing tank and mixing manifold for the HPST used in this work. Liquid fuel was introduced by direct injection using a gas-tight syringe (Hamilton Co., model 1010 TLL SYR) into a 12.8-liter stainless-steel mixing tank maintained at 120 o C. Synthetic air (i.e. 21% O 2/79% N 2), 4% O 2/96% Ar mixture, Airgon (i.e. 21% O 2/79% Ar) mixture, argon and nitrogen were from Praxair (research grade). Engine fuels were all obtained from the end-user (for example, RP fuels were from Edwards Air Force Base, jet fuels from Wright-Patterson Air Force Base). Single-component fuel (decalin and 1-butanol) were all from Sigma Aldrich without further preparation. A test gas mixture of fuel and bath gas was prepared manometrically and stirred using a magnetically-driven vane assembly for at least 30 minutes prior to the experiments. What is unique in this work on engine fuels is that a great deal of care was taken to handle the mixture preparation and to monitor the test gas mixture, ensuring that the fuel molecule right behind the reflected shock wave was the same as the fuel obtained from the storage bottle. The 18

45 monitoring process consisted of three steps: complete-evaporation check, minimal-loss mixture preparation, and in situ laser absorption measurement. More details can be found below. Figure 2.6 Schematic of mixing tank and mixing manifold for HPST. The fuel flask was not used in current studies. The gas bottles were prepared based on the experiment needs (i.e., not limited to 21% O 2/Ar and N 2). First, it is vital to make sure that the liquid fuel is fully evaporated inside the mixing tank; otherwise, the heavier components would be lost, introducing error in the pressure measurement of gas-phase fuel vapor and hence the preparation of the desired mixture. Throughout the whole course of engine fuel investigations these years, two methods have been adopted to check the completeness of evaporation. The first one is the use of an injection curve [75], i.e. the fuel pressure in a heated vessel (mixing tank here) as a function of injected liquid volume, to quantitatively establish the maximum allowed liquid volume for full evaporation, as shown in Figure 2.7. The assumption of doing this injection curve is that if the evaporation is complete, the fuel pressure should be proportional to the volume of injected liquid fuel; if not, the cut-off volume must be where the complete evaporation of liquid fuel starts to fail, indicating an upperlimit of injection volume that can be used in real shock tube experiments. As seen in Figure 2.7, the volume-pressure proportions remain good for A2 and C5 till 2.5 ml, and for C1 till 3 ml, while 5 ml A2 might be too much to guarantee full evaporation of A2 liquid. In real experiments, it was found that 2-3 ml was safe and adequate for all engine fuel tests (including both ignition and pyrolysis studies) [75]. The second way of checking evaporation is the use of a transparent flask (Chemglass CG , not shown) at 120 o C to ensure at least visually complete evaporation (i.e., no observable remaining droplets) before being introduced into the mixing tank. This qualitative check of fuel evaporation was no longer needed once confidence 19

46 was gained in the safe range of liquid volume for complete evaporation, and the first quantitative method was more routinely employed in this work A2 C1 C5 Pressure (Torr) T = 120 o C Volume (ml) Figure 2.7 Injection Curve of A2, C1, and C5. The vapor pressure of A2 is 140 Torr at 120 o C. Second, special caution was exercised whenever making a mixture of engine fuel and some bath gas. More concretely, after fuel evaporation in the mixing tank, the bath gas began to be filled into the tank to mix with the gas-phase fuel vapor. During this bath gas filling, the bath gas pressure in the mixing manifold was carefully adjusted to start with a pressure slightly above the fuel pressure in the tank (i.e. to avoid fuel loss at the beginning of filling), and was gradually controlled to maintain a slow and smooth filling rate (normally 1-2 Torr per second with no strong in-flow) [131]. All these controls were to minimize the possible loss of fuel components due to gas compression during the filling process. In addition, the length of mixing time was varied to see if there existed a minimum mixing time required for good homogeneity, but it was found that a mixing time from 15 mins to 2 hours made no difference in the fuel concentration measured by laser absorption and in the final kinetics targets [44, 96]. Lastly, it was always the in situ laser-absorption-measured value of fuel concentration in Region 1 (and often also in Region 5) that was used in the report of mixture composition in a reflected shock wave experiment of engine fuels. This is in big contrast to numerous studies of engine fuels where only the manometric concentration is used (as laser absorption measurements were not available) [8, 135]. It seems apparent that the fate of test fuel can be viewed in three phases: liquid in the storage container (Phase 1), gas in the mixing tank (Phase 2), and gas in the shock tube (Phase 3). Throughout these three phases, the fuel concentration is monitored such that a 20

47 good knowledge of fuel loss can be acquired. From Phase 1 to Phase 2, one can find how much fuel is recovered in the gas phase after being introduced from the storage container to the mixing tank (assuming the density of the liquid is known, and the ideal gas law is still valid), while from Phase 2 to Phase 3, one might figure out how much fuel is lost to the wall in transfer to the tube (i.e., the ratio of the Beer s Law concentration (initial one in Region 1) from in situ laser absorption measurement to the manometric concentration when the mixture is made). Table 2.2 shows the results of this fuel loss analysis for A2, C1, and C5 upon scrutiny of the author s experiment notes. The above 100% recovery in Phase 2 of C1 and C5 was possibly due to uncertainties in liquid volume reading, liquid density measurement, and tank temperature measurement, but in general, there was nearly no loss from liquid to gas in these experiments. However, the recovery in Phase 3 from Phase 2 varied with fuel identity, temperature of the tube (T 1), wall surface roughness, etc.. For A2, while only 80% on average was recovered at T 1 = 60 o C, it was raised to 95% by increasing T 1 to 90 o C. It is generally believed that, in the case of A2, condensation might be the main factor for fuel loss, and the tube should be hot enough to minimize this kind of fuel loss. Furthermore, it was found empirically that the partial pressure of fuel in the test gas mixture in the tube must lie below half or even one third of the saturated vapor pressure at the tube temperature [51]. C1 appeared to exhibit good recovery at both 90 o C and 110 o C, but C5 did not at either temperature. It is generally thought that, in the case of C5, adsorption might be the main factor for fuel loss. In other words, C5 might be inclined to stick to the wall however hot the tube would be. One way of reducing adsorption-induced fuel loss was passivation, i.e. forming a thin film of large fuel molecule on the wall to prevent further adsorption. In practice, in the case of C5 and other fuels of this behavior, either neat fuel or a fuel-rich mixture was used beforehand to rinse the wall [107]. In addition, it was found that the first shock from a single mixture usually showed the largest fuel loss, and the latter shocks had less loss. Therefore, only the shocks with acceptable (no more than 10%) fuel loss were used for data analysis and report. Table 2.2 Fuel loss analysis for A2, C1, and C5 Fuel Recovery in Phase 2 from Phase 1 Recovery in Phase 3 from Phase 2 (%, avg.) (%, avg.) A at 60 o C, 95 at 90 o C C at 90 o C, 110 o C C at 90 o C, 110 o C 21

48 22

49 3 Chapter 3 Ignition Studies of Engine Fuels 3.1 Example Measurements Example ignition experiments are shown in Figure 3.1. Data include pressure traces for the reactive (fuel/oxidizer) cases together with the OH* emission record and 3.39 μm He-Ne absorbance measurement. In this study, ignition delay time is defined as the time interval between time zero and onset of ignition. Time zero can be inferred from arrival of the reflected shock at the measurement location that manifests itself as a rapid rise in the pressure signal and the He-Ne absorbance. Onset of ignition can be deduced from the rising (pressure, OH* emission) or plummeting (He-Ne absorbance) of signal by extrapolating the maximum slope back to the corresponding baseline. Using this definition, three ignition delay readings are marked in Figure 3.1. For consistency, the ignition delay data discussed in the following sections use the emission definition unless stated otherwise. The He-Ne laser measurement tracks the decay of the parent fuel during the chemical induction process and its complete removal at ignition, which is consistent with the pressure and emission behaviors. Note our use of non-reactive pressure traces to ensure that the facility provides constant flow properties in the absence of reaction. It is worth noting that these traces, especially the pressure one, tell us more about the ignition process than just the delay time. In Figure 3.1, there is a more-evident larger pressure rise before the final ignition in the bottom panel than that in the top panel. This pressure rise is indicative of a homogeneous or inhomogeneous oxidative process with mild heat release typically known as pre-ignition. The reasons for this rise remain unclear, but there are ways (i.e. use of a constrained-reaction-volume) to circumvent it [128, 136, 137]. It is worth mentioning that this pre-ignition phenomenon typically occurs at times longer than 1 ms and/or temperatures near or below 1000 K in current work, and all current measurements are highly reproducible. Uncertainty of current ignition delay data is typically ±15-20%. This is estimated by the theory of propagation of uncertainty with the primary contribution from ±1% in the initial reflectedshock temperature. With activation energies for the majority of the experiments performed being near 40 kcal/mol, uncertainties of 1% in T 5, the reflected shock temperature, translates into an absolute uncertainty in the ignition delay times of ±15%. However, systematic differences in the data of 10% (e.g. because of pressure or fuel/oxygen loading) can readily be distinguished. 23

50 Pressure and He-Ne signal Pressure (reactive) 306 nm OH* emission 3.39 m He-Ne absorbance Pressure (non-reactive) A K 15.3 atm = 1.0 Bath gas: 4%O 2 /Ar t ign, pressure t ign, emission Time ( s) Emission signal 6 4 Pressure (reactive) 306 nm OH* emission 3.39 m He-Ne absorbance Pressure (non-reactive) t ign, He-Ne Signal 2 A K 11.1 atm = 1.0 Bath gas: Air t ign, pressure 0 t ign, emission Time ( s) Figure 3.1 Ignition delay time measurements using pressure (in black), 306 nm OH* emission (in red), and 3.39 µm He-Ne absorbance (in blue) during A2 oxidation. Initial reflected shock conditions: 1278 K, 15.3 atm, 4%O 2/Ar, equivalence ratio (ϕ) equal to 1.0 for top panel; 1031 K, 11.1 atm, ϕ = 1.0 for bottom panel. The non-reactive pressure profiles at similar conditions are also shown (in green). 3.2 Data and Correlations The correlations of the data are based on recent ignition delay time measurements in a variety of fuels. A brief summary of these fuels and their properties is given in Table 3.1 (for conventional fuels) and in Table 3.2 (for alternative fuels). A table of raw ignition delay times can be found in Appendix A1. 24

51 When analyzing the data and evaluating expressions for correlations, the data were binned into several groups, i.e. of high pressure (HP), low pressure (LP), Air versus O 2/Ar, variable equivalence ratio (ϕ), and variable O 2 mole fraction (X O2), as listed below. 1. HP Air IDT Data: distillate fuels in air at high pressures (10-60 atm) with ϕ = 1 ( ) 2. Previous HP Air IDT Data: previously published IDT data for distillate fuels in air at high pressure (10-60 atm) with ϕ = 1 ( ) 3. LP Air IDT Data: IDT data for distillate fuels in air at low pressure (2-6 atm) with ϕ = 1 4. HP 4% O 2/Ar IDT Data: distillate fuels in 4% O 2 in argon at high pressure (10-60 atm) with ϕ = 1 ( ) 5. LP 4% O 2/Ar IDT Data: distillate fuels in 4% O 2 in argon at low pressure (1-15 atm) with ϕ = 1 ( ) 6. HP N 2 vs Ar IDT Data: comparison of IDT of high pressure (12 atm) distillate fuels in Air and 21%O 2/Ar (Airgon) 7. Variable X O2 IDT Data: comparison of IDT with varying X O2, O 2 concentration 8. Variable ϕ IDT Data: comparison of IDT with varying ϕ In each section, a short history of the evolution of data is narrated, and the implications of this evolution on model development are discussed. 25

52 Table 3.1 Summary of ignition delay time data of neat conventional fuels. FUEL/Identifier P Range Molecular LHV a Aromatic POSF # (atm) Formula (MJ/kg) % Reference Data JP-8 (A1) C 10.8H this work Jet A (A2) C 11.4H this work JP-5 (A3) C 12.0H this work RP-2 (R4) C 12.0H this work RP-2 (R5) C 12.6H this work Kerosene (K6) C 11.5H 25.0 this work Kerosene (K7) C 12.0H 23.0 this work Kerosene (K8) C 13.7H 23.0 this work DF-2 (D9) C 13.6H [138] Gasoline (G10) C 7.4/H [138] Data before 2014 JP-8 (J11) C 10.9H [44, 64] Jet A (J12) C 10.17H [44] JP-8 (J13) C 11H 21 est [44] Jet A (J14) C 11H 22 est. [44] F-76 (F15) C 14.8H 26.9 [64] DF-2 (D16) C 12H 23 est. [65] Gasoline (G17) C 7.5H 13.9 est. 30 est. [70] Jet A (J18) C 11H 22 est [71] Jet A (J19) C 10.17H [8, 46] Kerosene (K20) C 9.49H [72] F-76 (F21) C 14.0H [74] a LHV: lower heating value 26

53 Table 3.2 Summary of ignition delay time data of alternative fuels. Not all fuel blends are included. FUEL/Identifier P Range Molecular LHV a Aromatic POSF # (atm) Formula (MJ/kg) % Reference Data Gevo (C1) C 12.5H this work C14/TMB (C2) C 12.3H this work A3/TMD (C3) C 12.8H this work Sasol/Gevo (C4) C 11.4H this work C10/TMB (C5) C 9.7H this work Virent (C6) C 11.9H this work Data before 2014 Gevo (C7) C 12.2H [64] Sasol IPK (S8) C 10.5H [64] Shell SPK (S9) C 9.6H [64] HRJ-tallow (H10) C 11.3H [64] HRJ-camelina C 11.6H [64] (H11) Swedish (B12) C 11.0H [64] F-76/HRA (H13) C 14.6H 29.5 [64] DIM (B14) C 13.4H 21.4 [64] (B15) C 13.6H 21.4 [64] a LHV: lower heating value HP Air IDT Data Conventional Fuels All current new ignition delay time (IDT) data in synthetic air (i.e. 21% O 2/79% N 2) for distillate fuels (A1 through G10 in Table 3.1) for equivalence ratios between 0.85 and 1.15 and for temperatures above 1000K and pressures from atm can be correlated with the simple formula t ign = P exp(15380/t) Eqn. 1 where t ign is in µs, P is in atm and the activation energy and temperature are in Kelvins. The same units are also used in Eqns Experiments were binned to include equivalence ratios 27

54 between 0.85 and 1.15, as over this temperature range and equivalence ratio, little sensitivity of t ign was seen in the IDT data (see later sections for a discussion of this point). The Standard Error for this correlation 1σ = 18% and for the activation energy is 480 K. The new data correlated with this expression are shown in Figure 3.2. Of particular note is that the IDT for the kerosene fuels (A1 - K8), gasoline (G10) and diesel fuel (D9) exemplars all correlate well with this expression, indicating little sensitivity to variations in the specific fuel compositions. Note that these fuels span molecular formulae from C 7.4H 13.7 (G10) to C 13.6H 25.9 (D9) and contain between 0.3% (R4) and 30.7% (G10) aromatic components (see Table 3.1). In fact, a similar observation was made in a controlled study on the effect of toluene on the autoignition of isooctane and n-heptane by Hartmann et al. [139] where they found, through both experiments and simulations, that the IDT of iso-octane/toluene or n-heptane/toluene mixture does not show a strong dependence on the toluene concentration for toluene concentrations less than 40% and temperatures above 1000 K. These similarities might lie in the observation that the pyrolysis product distributions of conventional fuels are similar (see Chapter 4), and might lie in the predominant role of two important reactions at high temperature, i.e. H + O 2 = OH + O and CO + OH = CO 2 + H. Since this role is relatively independent of the parent fuel composition, similar pyrolysis fragments are likely to lead to similar IDT behaviors at high temperature. Similarly, the apparent activation energy, 128 kj/mol found through Eqn. 1, might also be related to the pyrolysis system of conventional fuels. t ign ( s) A1 A2 A3 R4 R5 K6 K7 K8 D9 G K 1111K 1000K E A =128 kj/mol All normalized to 12 atm /T (1/K) Figure 3.2 Ignition delay time data for distillate fuel/air, ϕ = , normalized to 12 atm using the pressure dependence of Eqn. 1. The Eqn. 1 correlation is shown as a solid line. 28

55 Alternative Fuels All current new ignition delay time (IDT) data in synthetic air for C1 in Table 3.2 for equivalence ratios between 0.4 and 1.2 and for temperatures above 1000K and pressures from atm can be correlated with the simple formula t ign = P ϕ exp(16145/t) Eqn. 2 Evidently, ignition delay of C1 in air exhibits negative pressure-dependence (-1.156, close to - 1 as well) and negative yet weak equivalence-ratio-dependence (-0.356). Therefore, it is assumed P -1 for ignition delay times of C1 - C6 in the discussions below K 1111 K 1000 K C t ign ( s) 100 E A = 134 kj/mol All normalized to 12 atm /T (1/K) Figure 3.3 Ignition delay time data for C1/air, ϕ = , normalized to 12 atm using the scalings of Eqn. 2. The Eqn. 2 correlation is shown as a solid line. Figure 3.4 compares ignition delays of all C1 - C6 fuels in air at 12 atm and ϕ equal to 1.0 above 1000 K. First, it appears that these ignition delay data exhibit greater scatter, depending on the fuel identity, than that shown in Figure 3.2, and hence a common correlation for all C1 - C6 might not be possible. Second, while showing similar activation energy (within 5%) compared to the conventional fuels A1 - G10, these six alternative fuels are not all interchangeable with the conventional fuels in terms of ignition delay. More specifically, it appears that the ignition delays of C6 are close to those of A1 - G10, C5 and C2 are less reactive (by up to ~50%), while C1, C3, and C4 are more reactive (by up to ~20%). Nonetheless, caution must be exercised when generalizing such observations, as these observations are limited to 12 atm, K, and ϕ = 1.0. Comparisons at more comprehensive conditions might be needed, as will be shown in Section

56 1250 K 1111 K 1000 K t ign ( s) 1000 C1 C2 C3 C4 C5 C6 Eqn All normalized to 12 atm /T (1/K) Figure 3.4 Ignition delay times in air of six Category-C fuels at 12 atm and ϕ = 1.0 (solid dots). Dashed line is the correlation Eqn Evolution of Data To ensure reliability and reproducibility of our data, the author s A1 - A3 experiments in this section have been repeated by peer students, as shown in Figure 3.5. Clearly, the 2016 data by the author s peers are in excellent agreement with the author s own 2015 data, proving the good reproducibility of our measurements K 1111 K 1000 K 1250 K 1111 K 1000 K 1250 K 1111 K 1000 K 1000 A1 / Air Presure scaled to 12 atm ~ A2 / Air Presure scaled to 12 atm ~ A3 / Air Presure scaled to 12 atm ~ 1.0 t ign ( s) t ign ( s) t ign ( s) 100 A1/Air (2015) A1/Air (2016) Eqn /T (1/K) 100 A2/Air (2015) A2/Air (2016) Eqn /T (1/K) 100 A3/Air (2015) A3/Air (2016) Eqn /T (1/K) Figure 3.5 Reproducibility of current A1 - A3 HP Air IDT measurements. Additionally, the author s data have been re-analyzed every time the thermochemistry of test fuels is updated, thereby ensuring the correctness of shock condition calculations and allowing for reasonable comparisons between fuels. For example, during the C1 - C6 campaign, the fuel thermochemistry was updated three times (in March, April, and July 2015, respectively, see ref [58]). Since some C1 experiments were carried out even before the first set of thermochemical 30

57 data became available, an old thermo, namely, general conventional jet fuel (Jet-A) thermochemistry, was used for analysis of these C1 experiments at the very beginning. However, the new thermo yields non-negligibly different reflected shock conditions for C1 shocks from those using the old thermo. The reflected shock temperature (T 5) can differ by up to approximately 20 K, as shown in Figure 3.6. Here, the new thermo refers to any one of the thermochemical data mentioned above, since it is found that those three sets of thermochemical data yield minimal difference between each other (T 5 difference for C1 - C6: C1 < 1 K, C2 < 6 K, C3 < 3 K, C4 < 1 K, C5 < 2 K, C6 < 1 K). The correlation Eqn. 1 is also included in Figure 3.6. Clearly, C1 data reported with the old thermo appear nearly identical to the conventional fuel data, while the correct report of C1 data using the new thermo are actually around 20% lower. Therefore, again, care should be taken when comparing ignition delay times of conventional and alternative fuels, as assumptions about the thermochemistry of these fuels can significantly affect the temperatures determined in reflected shock wave experiments K 1111 K 1000 K C1 new thermo C1 old thermo Eqn. 1 t ign ( s) 100 Air 12 atm = /T (1/K) Figure 3.6 Influence of thermochemistry on C1 data report Previous HP Air IDT Data Previous conventional distillate fuel/air shock tube ignition delay time data (J11 - J14, J18, J19, G17) from the author s laboratory are compared with the correlation Eqn. 1 in Figure 3.7. These data exhibit greater scatter, but are still in reasonable agreement with Eqn. 1. Notably, earlier measurements of the same fuel (J19) by Wang and Oehlschlaeger [8, 46] and (J12) by Vasu et al. [44] are in good agreement with this correlation. Good agreement is also seen with the F-76 data (F21) from Gowdagiri et al. [74]. The largest deviation is seen in the shortest-time ignition 31

58 delay measurements (J19) by Dean et al. [71] that occurred at the highest temperatures (above 1200 K), suggesting the possible need to modify Eqn. 1 above 1200 K t ign ( s) K 1111K 1000K J11 J12 J13 J14 G17 J18 J19 F21 Eqn. 1 All normalized to 12 atm /T (1/K) Figure 3.7 Ignition delay time data for previously published fuel/air data, ϕ = , normalized to 12 atm using pressure dependence of Eqn. 1. The Eqn. 1 correlation is shown as a solid line LP Air IDT JP-8 Figure 3.8 shows ignition delay times of J11 in air at relatively lower pressures ( atm) and at ϕ in the range of Similarly, this set of data can be correlated with the simple expression t ign = P ϕ exp(16182/t) Eqn. 3 Comparing Eqns. 1 and 3 (and recognizing the weak equivalence ratio dependence), it is noted that the HP and LP ignition delays in air of conventional jet fuels share similar activation energy (within 5%) and pressure dependence (close to -1). Therefore, again, it is assumed P -1 for ignition delay times of alternative fuels shown below. 32

59 1429 K 1250 K 1111 K J11 (JP ) 1000 t ign ( s) 100 E A = 135 kj/mol All normalized to 3 atm /T (1/K) Figure 3.8 Ignition delay time data for J11/air, ϕ = , normalized to 3 atm using the scalings of Eqn. 3. The Eqn. 3 correlation is shown as a solid line Alternative Aviation Fuels Figure 3.9 and Figure 3.10 compare ignition delay data of J11 and some alternative aviation fuels at 3 and 6 atm, respectively. At 3 atm, both C7 and B12 are less reactive than J11, with C7 being the least reactive at relatively higher temperatures (above 1250 K). In addition, while B12 and J11 exhibit similar activation energy (134 and 135 kj/mol, respectively), C7 has a slightly lower one (110 kj/mol). At 6 atm, interestingly, Figure 3.10 shows that the ignition delay of conventional fuels agrees well with the data pool of alternative fuels for test. This observation is also seen in Figure 3.4 above. More specifically, while B12 is clearly the least reactive and H10 the most reactive at 6 atm, the ignition delay of J11 and other alternative fuels are similar. This finding might indicate that most alternative fuels, regardless of the manufacturing processes and individual components, are generally interchangeable with conventional fuels. Although this finding is consistent with previous observations for Jet-A and some Fischer- Tropsch alternative fuels at above 1000 K [8, 44], care should be emphasized when using this information for the application and deployment of these fuels. 33

60 1429 K 1250 K 1111 K 1000 C7 B12 Eqn. 3 t ign ( s) All normalized to 3 atm /T (1/K) Figure 3.9 Comparison of ignition delay time for J11 (JP , shown in dashed line), C7 (Gevo ATJ, shown in red dots), and B12 (Swedish BioJet, shown in blue dots) in air at 3 atm and ϕ = K 1250 K 1111 K 1000 K C7 H11 H10 S8 S9 B12 Eqn. 3 t ign ( s) 100 All normalized to 6 atm /T (1/K) Figure 3.10 Comparison of ignition delay time for J11 (JP ) and six alternative aviation fuels in air at 6 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn. 3. Additionally, the ignition delays of an example fuel blend of 50% conventional fuel and 50% alternative fuel are shown in Figure As expected, the ignition delay data of 50% B12 (Swedish BioJet)/50% J11 (JP-8) lie between those of B12 and those of J11 with nearly the same activation energy (~133 kj/mol). Another example is shown in Figure 3.12 where the ignition delay of 50% S9/50% J11 is nearly the same as those of S9 or those of J11 itself. These findings resonate with the drop-in requirement of alternative fuel strategies [3, 4], as blending 34

61 conventional and alternative fuels allows us to tailor alternative fuels to expected physical and chemical characteristics K 1250 K 1111 K % B12 / 50% J11 100% B12 Eqn. 3 t ign ( s) All normalized to 3 atm /T (1/K) 0.9 Figure 3.11 Comparison of ignition delay time for J11, Swedish BioJet, and 50/50 blend of J11 and Swedish BioJet in air at 3 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn K 1111 K 100% S9 50% S9 / 50% J11 Eqn t ign ( s) 100 All normalized to 6 atm /T (1/K) Figure 3.12 Comparison of ignition delay time for J11, SHELL SPK, and 50/50 blend of J11 and SHELL SPK in air at 6 atm and ϕ = 1.0. Ignition delay of J11 is shown by the correlation Eqn Evolution of Data As a preliminary effort toward understanding the differences between the selected alternative jet fuel (Gevo, C1) and the representative conventional jet fuel (Jet-A, A2), all relevant data collected in 2013 and 2015 are put together in Figure 3.13, with the same bath gas (air) and the 35

62 same equivalence ratio (ϕ = 1.0), but with various pressures (3, 6, 12, and 35 atm). The data at 3 and 6 atm are from Section 3.2.3, and those at 12 and 35 atm are from Section It should be noted that the Gevo and jet fuels studied in Section are not of the same POSF label as the ones in current section. For example, Gevo with POSF is the target in LP studies, while Gevo with POSF is the fuel in HP studies. Such information is detailed in Figure Understandably, this might affect the integrity of data comparison, but it should still be reasonable to assume that this disparity in POSF is of little influence in the corresponding ignition delay, given that these two Gevo batches are of very similar physical and chemical specifications [10, 58]. The same assumption is more likely to be valid for jet fuels (POSF 6169 in LP work, while POSF and in HP work), given that conventional jet fuels typically exhibit close ignition delay at high temperatures (see Figure 3.2). Therefore, it is assumed in Figure 3.13 that Gevo and represent the same fuel, Gevo, and jet fuel 6169, 10325, and refer to the same fuel, JP-8. Several interesting observations can be made through this comparison. First, it appears that there always exists a cross-over in ignition delay time between Gevo and JP-8, with varying cross-over temperature with the pressure. More specifically, the cross-over temperature is approximately 1200 K at 3 and 6 atm, while it is approximately 1050 K at 12 and 35 atm. Second, evidently, the relative magnitude of ignition delay between Gevo and JP-8 evolves around the cross-over temperature at a given pressure. More concretely, within the temperature window ( K) at 3 or 6 atm, or at temperatures below 1050 K and at 35 atm, Gevo appears to ignite slower than JP-8, while within the window of K and at 6 or 12 atm, Gevo appears to ignite faster than JP-8. Admittedly, the more data acquired within these temperature windows, the more conclusive these findings can be. Nonetheless, these comparisons indicate that how Gevo compares to JP-8 depends on temperature and pressure, and hence it may be incorrect to claim that Gevo must always ignite slower than JP-8 simply due to considerable amount of iso-paraffinic constituents. 36

63 t ign ( s) K 1250 K 1111 K 1000 K 909 K Fuel/Air = 1 3 and 6 atm: Gevo and JP atm: Gevo and JP atm: Gevo and JP atm 12 atm 6 atm 35 atm 100 JP-8 Gevo Lines: best fits /T (1/K) Figure 3.13 Comparison of JP-8 and Gevo Air IDT at various pressures HP 4% O2/Ar IDT Data Conventional Fuels Figure 3.14 includes high pressure (10-16 atm and atm) conventional fuel data measured in 4% O 2/balance argon (A1, A2, R4, R5, D9, J11, F15). These data (with equivalence ratios between 0.85 and 1.15) can be correlated with the similar formula t ign = P exp(17530/t) Eqn. 4 Notably, the ignition delay time data in argon exhibit smaller scatter (Standard Error is 14%) as shock wave experiments in a monatomic carrier gas do not experience large bifurcation in the reflected shock regime or suffer significantly from possible vibrational relaxation effects [130]. Also notable is that the activation energy (Standard Error = 340 K) of the lower concentration 4% O 2/balance argon experiments is slightly larger (by around 15%) than the activation energy for the fuel/air data shown in Figure 3.2. Both sets of earlier (2013) data (J11 & F15) are in excellent agreement with the correlation [44, 64]. 37

64 t ign ( s) A1 A2 R4 R5 D9 J11 F K 1250K 1111K 1000K E A =146 kj/mol All normalized to 12 atm /T (1/K) Figure 3.14 Ignition delay time data for fuel/4% O 2/Ar data, ϕ = , normalized to 12 atm using pressure scaling of Eqn. 4. The Eqn. 4 correlation is shown as a solid line. Earlier data in 2013 are shown with open symbols Alternative Fuels Comparison of conventional and alternative jet fuels can be further made in HP 4%O 2/Ar, as shown in Figure Figure 3.15 shows a comparison of A2 (representative conventional jet fuel), C1, and C5 (C1 and C5 are selected alternative jet fuels by FAA) in two kinds of bath gas, i.e. 4%O 2/Ar and air, at 12 atm and ϕ = 1.0. Most noticeably, while C1 ignites slower than A2 approximately by a factor of 2 in 4% O 2/Ar, it ignites faster than A2 by around 20% in air. However, unlike C1, C5 is consistently less reactive (by around 25%) than A2 in both bath gases. The mechanistic reasons remain unclear for the inconsistent behaviors of C1 and A2 at current conditions. Many factors, including oxygen concentration and temperature range effects (see Figure 3.13), Ar vs. N 2 effects, etc., are likely to be relevant in explaining the differences between conventional and alternative fuels. One hypothesis lies in the finding that the pyrolysis product distributions of C1 and A2 are significantly different (see Sections 4.1 and 4.2). The major pyrolysis product of A2 is ethylene, while that of C1 is iso-butene [140], and hence it is likely that the fuel chemistry differences between ethylene and iso-butene lead to the different IDT behaviors between A2 and C1 discussed above. For instance, Figure 3.16 shows kinetic modeling of IDT of ethylene and iso-butene in two kinds of bath gas at 12 atm and ϕ = 1.0 using the USC Mech II [141] and constant UV assumption. The bath gas (oxygen concentration) effect on iso-butene IDT appears more significant than that on ethylene IDT, and hence it is likely to lead to the more dramatic difference between 4%O 2/Ar IDT and Air IDT for C1 than that for 38

65 A2 as seen in Figure More experimental and theoretical investigations of IDT of ethylene and iso-butene at similar conditions in Figure 3.15 might be useful in supporting this hypothesis K 1250 K 1111 K 1000 K 4%O 2 /Ar t ign ( s) 1000 C1 Air 100 C5 A2 C5 A2 C1 FAA fuels 12 atm = 1.0 scaled data Lines: best fits /T (1/K) Figure 3.15 Comparison of measured IDT of A2 (in green), C1 (in black), and C5 (in red) in two kinds of bath gas, i.e. 4% O 2/Ar (hollow dot with a cross) and air (solid dot), at 12 atm and ϕ = 1.0. Solid lines are best fits to data points K 1429 K 1250 K 1111 K ic 4 H 8 4%O 2 /Ar ic 4 H 8 Air t ign ( s) 100 C 2 H 4 4%O 2 /Ar C 2 H 4 Air 12 atm = 1.0 USC Mech II Const. UV /T (1/K) Figure 3.16 Comparison of simulated IDT of ethylene (in green) and iso-butene (in black) in two kinds of bath gas, i.e. 4% O 2/Ar (hollow dot with a cross) and air (solid dot), at 12 atm and ϕ = 1.0. Solid lines are best fits. Simulations done using USC Mech II [141] and constant UV assumption. A comparison of the measured ignition delay times for F15 (F-76 for navy applications) and some alternative navy fuels is shown in Figure 3.17 for a pressure of 17 atm and an equivalence ratio of 1.0. The low scatter of the data shown in Figure 3.17 permits us to distinguish small 39

66 differences in ignition delay of the different fuels. Generally, there is fair agreement between the ignition delay times and activation energies for these four fuels. Specifically, ignition delay times of F15 appear to be approximately 40% longer than those of H13 at above 1200 K, but they tend to converge at lower temperatures. In contrast, ignition delay times of B14 are consistently longer (~20%) than those of B15. However, again, owing to the possible variability of the composition of the hydro-refined algal oil and terpene dimers used to blend these alternative fuels (H13, B14, B15), caution must be exercised in generalizing these findings K 1250 K 1111 K F15 H B14 B15 Eqn. 4 t ign ( s) 100 All normalized to 17 atm /T (1/K) Figure 3.17 Comparison of measured IDT for F15 (F-76), H13 (F-76/HRA Blend), B14 (DIM), and B15 (103-2) in 4%O 2/Ar at 17 atm and ϕ = 1.0. Correlation using Eqn. 4 is shown in dashed line HP N2 vs Ar IDT Data There is an interest in comparing HP IDT in 21%O 2/N 2 (Air) and 21%O 2/Ar ( Airgon for brevity). Haylett et al. found only small variations of a few percent in the ignition delay times of n-dodecane/air compared to n-dodecane/airgon [65]. Conversely, Shen et al. observed that the ignition delays of iso-octane in Airgon are 20% - 40% shorter than those in Air [142]. Having recognized the possible controversy in this comparison, the author and a peer student performed the same experiments in 2015 and 2016, respectively, as listed in Table

67 Table 3.3 Datasets of Category-A fuels in Air and Airgon FUEL/Identifier Bath Person Year JP-8 (A1) Air YZ, JK 2015, 2016 Jet A (A2) Air YZ, JK 2015, 2016 JP-5 (A3) Air YZ, JK 2015, 2016 JP-8 (A1) Airgon JK 2016 Jet A (A2) Airgon YZ, JK 2015, 2016 JP-5 (A3) Airgon JK 2016 As stated in Section , the Air experiments are highly reproducible. To see the repeatability of Airgon experiments, Figure 3.18 plots the author s and the peer s ignition delay data of A2 in Airgon, as well as our Air data of A2. All data in Figure 3.18 fall within the equivalence ratio range of , and are normalized to 12 atm using the pressure scaling of Eqn. 1. Evidently, the author s and the peer s data are in reasonably good agreement, and both sets of Airgon data generally appear shorter than the Air data. Specifically, the two sets of Airgon data at temperatures of K appear closer to the Air data than those Airgon data at above 1150 K. This might be attributed to the influence of bifurcation effects on the time zero determination in the Airgon (and Air) IDT experiments [130]. This influence tends to be bigger at the high temperature end where the ignition delay is shorter than 150 µs [143]. One might argue that because of the lower heat capacity (C p) for the monatomic carrier gas compared to the diatomic nitrogen carrier gas, temperatures modeled using either a constant P or constant V assumption can change differently and in different directions depending on whether the reaction chemistry is endothermic or exothermic. In the case of fuel ignition, during fuel decomposition, an endothermic process, the temperature drop with argon as the carrier gas is larger than in nitrogen. However, during the oxidation process which is exothermic, the temperature rise is larger in argon than in nitrogen. These two effects have an opposite effect on the overall ignition delay time. Argon lengthens the decomposition step, but shortens the oxidation step. The same changes happen with nitrogen, but to a lesser degree. However, more data, especially at above 1150 K, are needed in the future to draw a more compelling conclusion. 41

68 1250 K 1111 K 1000 K A2/Air A2/Airgon - JK A2/Airgon - YZ 1000 t ign ( s) atm ~ /T (1/K) Figure 3.18 A2 Air and Airgon IDT with ϕ = only. Aside from the bifurcation effect, current Air and Airgon data comparison might still be complicated by the fact that the measured pressure traces in Airgon experiments exhibit more conspicuous and relatively stronger induction period pressure rise, or pre-ignition mentioned above. For example, Figure 3.19 shows a comparison of pressure traces in Air and Airgon for the reflected shock experiment at ~1070 K, ~12 atm, and ϕ ~ 1.0. The pressure traces are, from left to right, the author s work in Air, the author s in Airgon, and the peer s in Airgon, respectively, with the pre-ignition segment shown in red. The pre-ignition segment can be characterized by how long it lasts within the total ignition delay and how smoothly it transitions to the final ignition pressure jump. Evidently, first, the author s and the peer s pre-ignition pressure signals in the Airgon experiment consistently dominate nearly the latter half of the entire induction time. Second, these signals are consistently slowly increasing over time by approximately 20% from baseline. Third, these signals are consistently followed by a sudden, nearly instantaneous jump, with hardly any sign of smooth and continuous transition. On the contrary, the pre-ignition signal in the Air experiment lasts for much shorter percentage period of time, and makes a much smoother transition to the final ignition pressure signal. These differences in the pre-ignition behavior might indicate that the Airgon experiment experiences a different thermodynamic/gasdynamic environment, thereby making it unfair to directly compare the corresponding IDT to the Air IDT with other conditions being equal. Therefore, again, it is suggested that more Airgon data be collected such that their pre-ignition behavior is better controlled. One feasible way of control is lower the fuel/oxidizer concentration, say 4%O 2/N 2, to reduce heat release. Another more advanced way of pressure control, i.e. CRV, 42

69 will be discussed in length in Chapter 6. It is highly possible that without the pre-ignition complication, the Air and Airgon IDT should effectively be identical. In fact, several kinetics simulations have indicated that this is generally true [81, 144]. Section 3.3 will show some example modeling results that might shed some light on this. 5 4 YZ A2/Air K 11.7 atm = YZ A2/Airgon K 15.1 atm = JK A2/Airgon K 16.6 atm = 1.12 Pressure Signal Pressure Signal Pressure Signal Time ( s) Time ( s) Time ( s) a b Figure 3.19 Pressure trace in A2 Air and Airgon IDT experiments. c Variable XO2 IDT Data Figure 3.20 shows a comparison of IDT values for jet fuel (A2) with variable O 2 concentration for equivalence ratios near unity. Note that the Air IDT data instead of Airgon IDT data are included in this plot, since the pressure behaviors in the Airgon experiments are different from those in the IDT experiments in other bath gases (see Figure 3.19 and Figure 3.1) (and it is assumed that the Airgon IDT data, without the pre-ignition complication discussed above, are the same as the Air IDT data). As seen in Figure 3.20, the variation in O 2 concentration scaling is temperature dependent. The activation energy appears to drop as the fuel and oxygen concentration are increased. In the case of the 1.2 atm 4% O 2/Ar experiments, these data appear to show a relatively higher activation energy than that at 12 atm. At 12 atm in Ar, the ratio of the ignition delay times for A2 fuel with 4% and 21% O 2 (a ratio of 1/5.25) implies an IDT scaling of X O2 in the overlapping temperature window. A simple correlation of IDT with O 2 concentration over the full concentration set, however, is not yet possible. 43

70 K 1250 K 1111 K 1000 K 500ppm Fuel/O 2 /Ar 1.2 atm 4%O 2 /Argon 1.2 atm 12 atm 21%O 2 /N 2 12 atm 1000 t ign ( s) /T (1/K) Figure 3.20 A comparison of A2 jet fuel ignition delay time data for ϕ = 1 for varying amounts of O 2 in Ar and N 2. The 500ppm fuel/o 2/Ar data is J11 POSF 6169, which is very similar to the A2 fuel, and has 0.82% O 2 for ϕ = Variable ϕ IDT Data The ignition delay time correlations in the sections above were limited to experiments with equivalence ratios between 0.85 and This is because an examination of a larger set of data with a wider range of equivalence ratios shows only limited variation of the IDT with equivalence ratio in this range (see Eqns. 2 and 3). As another example, Figure 3.21 shows the variation of ignition delay time with equivalence ratio for A1 jet fuel for two pressures: 1.2 and 12 atm. What is evident in Figure 3.21 is that the dependence of ignition delay time on equivalence ratio varies with temperature. In addition, this variation produces a cross-over point or temperature, near 1200 K, where there is effectively no dependence of ignition delay time on equivalence ratio for A2 jet fuel. Above 1200 K, as equivalence ratio is decreased, ignition delay times get shorter. Below 1200 K, as equivalence ratio is decreased, ignition delay times get longer. In either case, in this temperature range, small changes in equivalence ratio near values of unity, do not strongly affect the ignition delay times. Similarly, the equivalence ratio dependence of ignition delay times of F15 (F-76) in 4% O 2/Ar mixture is shown in Figure The activation energy at 17 atm decreases with increasing equivalence ratio with a cross-over temperature at around 1250 K. In contrast to this, however, Dean et al. [71] found in their study at atm that higher equivalence ratios always resulted in shorter IDT values, and no cross-over 44

71 temperature was observed in their data up to 1660 K. Further work is needed to characterize this cross-over behavior atm, A2/4% O 2 /Argon =2.0 =1.1 =1.0 = K 1111K 1000K t ign ( s) atm, A2/Air =1.0 = /T (1/K) Figure 3.21 Variation of ignition delay time with equivalence ratio for A1 jet fuel K 1250 K 1111 K 17 atm F11/4% O 2 /Ar 10 3 t ign ( s) 10 2 = 0.54 = 1.0 = /T (1/K) Figure 3.22 Variation of ignition delay time with equivalence ratio for F-76 in 4%O 2/Ar. 3.3 Modeling Experimental results are further compared to recent available predictions using two modeling approaches, i.e. the surrogate approach and the HyChem approach. 45

72 3.3.1 Surrogate Approach As mentioned earlier, both well-conceived surrogate compounds and widely validated reaction mechanisms are needed to make reasonable kinetic simulations for these practical fuels. On one hand, for JP-8, there exist various surrogate fuels, ranging from a simple two-component surrogate [145] to a complex twelve-component one [146]. In addition, the methodology of surrogates formulation has been evolving, from direct hydrocarbon-class compositional analysis [37], to a recent, more systematic way of using fundamental combustion property targets [10, 46, 49, 50], such as Derived Cetane Number (DCN), H/C ratio, Threshold Sooting Index (TSI), and so on, as listed in Table 3.4. However, this is not the case for alternative fuels since they have only emerged recently. Furthermore, less has been published about the hydrocarbon class compositions, sooting behaviors, and other general specifications of Category-C fuels, F15 and alternative navy fuels [6, 11], much less their corresponding surrogates. On the other hand, the availability of reaction mechanisms for predicting combustion kinetic phenomena of practical fuels relies on the selection of relevant surrogate fuels. Unfortunately, the lack or immaturity of reaction schemes for many specific surrogate components limits the immediate use of many reaction mechanisms. In this study, selection of kinetic models takes into consideration availability, validity, and consistency of both surrogate fuels and reaction mechanisms. As an example of this selection endeavor, Table 3.5 lists the qualitative findings of running several models of surrogate mixture and reaction mechanism combination for LP Air IDT of J11. The most satisfactory combination here, i.e. Dooley No.2 surrogate and Malewicki mechanism, was discovered merely through trial and error. After this type of model screening, we simply appraise the capabilities of several good multi-component surrogate/mechanism combinations and the implications this has on modeling experimental data. 46

73 Table 3.4 Some combustion property targets used in surrogate formulation for real fuels. Data were taken from [10, 46, 49, 50]. Name Derived Cetane Number (DCN) H/C Ratio Threshold Sooting Index (TSI) JP-8 (J11) SASOL IPK (S8) SHELL SPK (S9) HRJ-Tallow (H10) HRJ-Camelina (H11) GEVO ATJ (C7) Swedish BioJet (B12) N/A N/A F-76 (F15) N/A F-76 Blend (H13) N/A DIM (B14) N/A N/A (B15) N/A N/A Table 3.5 Screening of combination of surrogate compounds and reaction mechanism for surrogate modeling of J11 IDT (not an exhaustive effort). Lindstedt S2 [145] Violi #3 [26] Ranzi [82] JetSurf [144] Malewicki [53] underpredict LP Air IDT overprediction by a factor of 2 but similar action energy for LP Air IDT generally good for HP Air IDT except slightly earlier roll-off [44], underpredict LP Air IDT Violi #1 [26] Dooley #2 [49] no tetradecane submechanism no 1,3,5-trimethylbenzene submechanism good for LP Air IDT Table 3.6 describes the surrogate composition and reaction mechanism for four fuels being modeled. These predictions are also shown in Figure Figure The traditional constant energy and constant volume gasdynamic assumption is used in these predictions. All simulations were performed using the CHEMKIN-PRO software package [147]. 47

74 Table 3.6 Surrogate compositions and reaction mechanisms for fuels in this study Identifier Surrogate Composition (mol%) Reaction Mechanism J11 S9 F15 R6 [49]: 40.41% n-dodecane 29.48% iso-octane 7.28% 1,3,5-trimethylbenzene 22.83% n-propylbenzene [48]: 28.0% iso-octane 61.0% n-decane 11.0% n-dodecane [35]: 43.3% n-hexadecane 30.0% decahydronaphthalene 26.7% 1-methylnaphthalene [145]: 89% n-decane 11% toluene [53]: 2080 species, 8310 reactions [48]: 753 species, 4668 reactions [82]: 435 species, reactions [144]: JetSurf Aviation Fuels Figure 3.23 shows ignition delay times of JP-8 (J11) in air at various pressures (3 and 6 atm) and equivalence ratios (0.5, 1.0, and 1.5) and corresponding mechanism predictions. Note the good agreement between data and modeling at 3 and 6 atm, ϕ = 1.0, especially at 6 atm. However, while current J11 data yield a slightly positive equivalence ratio dependence, the modeling predicts a slightly negative dependence. 48

75 1429 K 1250 K 1111 K 1000 K 3 atm = t ign ( s) 3 atm = atm = atm = 1.0 JP-8/Air Dots: current work (scaled) Lines: surrogate modeling /T (1/K) Figure 3.23 J11 IDT in air. Data are shown in solid dots and kinetic modeling (see Table 3.6) in dashed lines. Additionally, the kinetic modeling of SHELL SPK (S9) is plotted in Figure Note, again, the good agreement between data and modeling K 1176 K 1053 K S t ign ( s) 50% S9 50% J11 Blend J11 Fuel/Air 6 atm = 1.0 Dots: current work (scaled) Lines: surrogate modeling /T (1/K) Figure 3.24 Comparison of IDT for JP-8, SHELL SPK, and 50/50 blend of JP-8 and SHELL SPK in air at 6 atm and ϕ = 1.0. Data are shown in solid dots and kinetic modeling (see Table 3.6) in dashed lines. The successful kinetic predictions of ignition delay for JP-8 (J11) and SHELL SPK (S9) imply the importance of a systematic approach in formulating the most appropriate surrogate fuel compounds for practical fuels, such as Dooley et al. s method of matching molecular weight, DCN, H/C ratio and TSI [49]. Information about these targets is available for some alternative fuels in Table 3.4 and will be helpful to the formulation of their surrogate fuels. Once the 49

76 surrogate components have been identified and agreed upon, efforts should be made on the development of chemical kinetic models for accurate representation of this surrogate [37]. For instance, the aforementioned inconsistency in equivalence ratio dependence of JP-8 between data and modeling suggests that Malewicki et al. s reaction mechanism [53] could be refined, and in particular, further experimental and theoretical studies for specific surrogate components, such as larger aromatics (e.g. 1,3,5-trimethylbenzene), may be needed Navy Fuels The variation behavior of F15 IDT with ϕ seen in Figure 3.22 appears confirmed by the mechanism predictions at those equivalence ratios and at 17 atm. Moreover, the kinetic modeling also appears to capture the pressure dependence of ignition delay of F-76, as indicated by the good agreement between data and modeling at both 17 and 44 atm, ϕ = K 1250 K 1111 K 10 3 F-76/4% O 2 /Ar Squares: current work (scaled) Lines: surrogate modeling t ign ( s) 10 2 P = 17 atm = 0.26 P = 17 atm = 0.54 P = 17 atm = 1.0 P = 17 atm = 1.2 P = 44 atm = /T (1/K) Figure 3.25 F15 IDT in 4%O 2/Ar. Data are shown in solid squares and kinetic modeling (see Table 3.6) in dashed lines. This success in kinetic modeling of F15 ignition is, again, due to the proper selection of surrogate composition [35] and widely validated reaction mechanism [82, 148] containing the surrogate fuel compounds. The surrogate composition chosen for F-76 was in fact designed for diesel fuel, but due to similarities in molecular weight [149] and DCN [6] between F-76 and diesel fuel, this surrogate holds a strong potential for representing the combustion behavior of F

77 RP Fuels Figure 3.26 shows measured and simulated IDT of RP-1 (R6) in 4%O 2/Ar at two pressures (LEFT) and two equivalence ratios (RIGHT). Note the goodness of a simple two-component surrogate fuel for predicting the IDT of RP-1. t ign ( s) K RP-1/4%O 2 /Ar =1.0 P K 1042 K P=17 atm P=43 atm t ign ( s) K 1190 K 1042 K RP-1/4%O 2 /Ar 17atm P =0.5 = Dashed line: 89%/11% Decane/Toluene JetSurf Dashed line: 89%/11% Decane/Toluene JetSurf /T (1/K) /T (1/K) Figure 3.26 RP-1 IDT in 4%O 2/Ar. Data are shown in solid squares and kinetic modeling (see Table 3.6) in dashed lines HyChem Approach Compared to the surrogate approach, the HyChem approach appears to be a more straightforward and predictable methodology. Figure show the modeling results of ignition delay time of conventional and alternative jet fuels using the HyChem approach, and Figure present the results of RP fuels. All HyChem simulations were run using the Version 2 of the HyChem mechanism unless stated otherwise (see Table 3.7, [150]) and the CHEMKIN-PRO software package. Overall, the experimental data and HyChem simulations are in good agreement. This appears remarkable in that no rate constant adjustment was attempted in matching the measured ignition delays during the development of the HyChem models [56, 69]. More specific merits and shortcomings of these simulations are discussed in detail below Jet Fuels Figure Figure 3.29 compare measured and simulated IDT in the bath gases of 4%O 2/Ar and air of A1, A2, and A3, respectively. Overall, the HyChem simulations for A1 - A3 are in reasonably good agreement with the measurements and qualitatively capture the corresponding equivalence ratio or pressure dependence of IDT. But there are still room for improvement. 51

78 For example, for A1, while the measured ignition delays at ϕ = 2.1 in 4%O 2/Ar appear close to those at ϕ = 1.1, the simulations do not reproduce this closeness; while the measured ignition delays at ϕ = 0.4 in air appear slightly longer (approximately by 15%) than those at ϕ = 1.0, the simulations do not reproduce this difference. Furthermore, it appears that for A2 and A3, the simulations in 4%O 2/Ar are slightly shorter (approximately by 15%) than the measurements. Additionally, at pressures higher than 14 atm, say, 54 atm for A1 in 4%O 2/Ar, 35 atm for A1 in air, and 32 atm for A3 in air, the disparity between data and modeling appears slightly larger (approximately by 20%) than that at atm K 1250 K 1111 K 1000 K 13 atm = atm = atm = atm = K 1111 K 1000 K 12 atm = atm = atm = 1.0 t ign ( s) 1000 t ign ( s) A1/4%O 2 /Ar Dots: current work (scaled) Lines: HyChem modeling /T (1/K) 10 A1/Air Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.27 A1 4%O 2/Ar and Air IDT data and HyChem simulations K 1250 K 1111 K 1000 K 14 atm = 1.0 4% O 2 /Ar 11 atm = 1.1 Air 1000 t ign ( s) A2 Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.28 A2 4%O 2/Ar and Air IDT data and HyChem simulations. 52

79 K 1250 K 1111 K 1000 K 14 atm = 1.0 4% O 2 /Ar 14 atm = 1.2 Air 32 atm = 1.1 Air 1000 t ign ( s) A3 Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.29 A3 4%O 2/Ar and Air IDT data and HyChem simulations. Similar data and HyChem modeling comparisons can be made for the two FAA-selected Category-C alternative fuels, C1 and C5, as shown in Figure 3.30 and Figure In general, again, the HyChem modeling does a good job in reproducing the IDT of C1 and C5 in both 4%O 2/Ar and air, and capturing the equivalence ratio and pressure dependences of C1 IDT in air. Nonetheless, it appears that the modeling for C1 IDT in 4%O 2/Ar need more refinement, since the simulations generally overpredict by approximately 15% and exhibit some curvature that is not seen in the measurements K 1250 K 1111 K 1000 K 15 atm = 1.0 4% O 2 /Ar 12 atm = 0.9 Air 11 atm = 0.4 Air 37 atm = 1.2 Air t ign ( s) C1 Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.30 C1 4%O 2/Ar and Air IDT data and HyChem simulations. 53

80 K 1250 K 1111 K 1000 K 15 atm = 1.0 4% O 2 /Ar 12 atm = 1.0 Air 1000 t ign ( s) C5 Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.31 C5 4%O 2/Ar and Air IDT data and HyChem simulations RP Fuels Lastly, the HyChem model is used to simulate the IDT of two RP-2 fuels, R4 and R5, in 4%O 2/Ar and air, in Figure 3.32 and Figure 3.33, respectively. The overall agreement between data and modeling is good, except that in the case of R5 IDT in air, the simulations appear to fail to capture the slight fall-off of data at near 1000 K. This might suggest that there is a need to refine the chemistry at intermediate-to-low temperatures K 1250 K 1111 K 1000 K 13 atm = 1.0 4% O 2 /Ar 11 atm = 1.0 Air 1000 t ign ( s) R4 (RP-2-1) Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.32 R4 4%O 2/Ar and Air IDT data and HyChem simulations. 54

81 K 1250 K 1111 K 1000 K 13 atm = 1.2 4% O 2 /Ar 11 atm = 1.0 Air 1000 t ign ( s) R5 (RP-2-2) Dots: current work (scaled) Lines: HyChem modeling /T (1/K) Figure 3.33 R5 4%O 2/Ar and Air IDT data and HyChem simulations Comparison of Surrogate and HyChem Simulations JP-8 IDT Simulations As depicted above, notably, it is JP-8 (A1, A2, A3, J11) only that enjoys the luxury of both surrogate model and HyChem model in current studies. Since the aforementioned JP-8 surrogate model (see Table 3.6) and the HyChem Category-A models all show equally satisfactory agreement with the data, it seems obvious that both surrogate and HyChem modeling approaches work well for JP-8 modeling, as shown by Figure 3.34 below (except that in the case of 4%O 2/Ar the HyChem simulations appear slightly better than the surrogate model simulations). However, it is worth mentioning that the current JP-8 surrogate model is the best available and carefully screened model. In other words, simulations using a surrogate model other than the current one would fail to match the performance of current JP-8 surrogate model or the HyChem model. On the contrary, it takes fewer screening efforts and faster iteration cycles using the HyChem approach to arrive at a model that is directly linked to a particular engine fuel, especially an alternative fuel whose properties and specifications might be beyond our experience base. Therefore, in short, the HyChem approach might be a better choice than the surrogate approach in the long run of development and refinement of combustion kinetics models of new engine fuels. 55

82 K 1250 K 1111 K 1000 K Data HyChem Surrogate K 1111 K 1000 K Data HyChem Surrogate t ign ( s) 1000 t ign ( s) A1/4%O 2 /Ar 13 atm = 1.1 Dots: current work (scaled) Lines: kinetic modeling /T (1/K) 10 A1/Air 12 atm = 1.0 Dots: current work (scaled) Lines: kinetic modeling /T (1/K) Figure 3.34 A1 IDT in 4%O 2/Ar and Air. Simulations using the HyChem approach are shown in Red and simulations using the surrogate approach (see Table 3.6, surrogate using [49] and mechanism using [53]) are shown in Black Air and Airgon IDT Revisited Figure 3.35 shows a comparison of simulated pressure trace in Air and Airgon during ignition of A2 at 1070 K, 12 atm, and ϕ = 1.0, using the surrogate approach (LEFT) and the HyChem approach (RIGHT), respectively. All simulations were carried out using the constant-energy, constant-volume assumption. In spite of the minor difference between the IDT in Air and in Airgon (and the final pressure level) in two plots, both reveal that the shapes of the pressure trace in Air and Airgon are much alike, including the pre-ignition part where the pressure trace first rises gradually and then switches smoothly up to the largest jump. This shape of pressure signal is similar to that in the Air experiment shown in Figure 3.19 a, but different from those in the Airgon experiments shown in Figure 3.19 b and c, indicating that the Airgon data presented in Section might not qualify for being modeled as a constant-energy, constantvolume reactor. Additionally, while both surrogate and HyChem modeling are consistent in the magnitude of the final pressure plateau and in the relatively higher pressure level in the Airgon case compared to the Air case, they are not consistent in the ignition delay, with the surrogate modeling showing shorter IDT and the HyChem modeling showing longer IDT in the Airgon case. This might be related to their differences in the way of modeling the temperature change effects during the oxidation process (see Section 3.2.5). Further work is needed to better understand these differences and to interpret our data. 56

83 50 Air Airgon 50 Air Airgon Pressure (atm) A2 Surrogate Model 1070 K 12 atm = 1.0 Pressure (atm) A2 HyChem Model 1070 K 12 atm = Time ( s) Time ( s) Figure 3.35 Comparison of simulated pressure traces in Air and Airgon using the surrogate approach (LEFT) and the HyChem approach (RIGHT) Evolution of the HyChem Mechanism The HyChem mechanism has been improving over time. Table 3.7 lists the release dates of Version 1.0 and Version 2.0 of the HyChem model for each of the seven fuels of interest to FAA, i.e. A1, A2, A3, C1, C5, R4, and R5 [150]. Through running simulations for current shock tube IDT experiments using both versions of the HyChem model, it is found that the C5 and R5 models have the least change in terms of IDT mechanism predictions. In addition, updates of the C1 model were mainly based on the author s peer s multi-species pyrolysis work that will not be included here. Hence, only the evolution of the Category-A HyChem models is briefly discussed here. Table 3.7 The HyChem model versions [150]. Fuel Model Version 1.0 Model Version 2.0 A1 (POSF10264) Mar. 28, 2015 May 24, 2016 A2 (POSF10325) Mar. 28, 2015 May 24, 2016 A3 (POSF10289) Mar. 28, 2015 May 24, 2016 C1 (POSF11498) Sept. 11, 2015 May 24, 2016 C5 (POSF12345) Aug. 27, 2015 Jun. 25, 2016 R4 (RP2-1 POSF7688) N/A Jun. 1, 2016 R5 (RP2-2 POSF5433) May 30, 2015 May 27,

84 The biggest change for the Category-A HyChem models lies in the significant improvement over the prediction of ignition delay time in air. Figure 3.36 gives an example. Evidently, Version 1 overpredicts the ignition delay of A2 in air by a factor of 3-4, while Version 2 predictions match the data reasonably well. The main reason for this is that Version 2 used an optimization method to determine the model parameters (stoichiometric coefficients and reaction rate coefficients) [ ] K 1111 K 1000 K Data HyChem V01 HyChem V02 t ign ( s) A2/Air 11 atm = 1.1 Dots: current work (scaled) Lines: kinetic modeling /T (1/K) Figure 3.36 Air IDT of A2 using version 1 and version 2 of the HyChem mechanism. 58

85 4 Chapter 4 Pyrolysis Studies of Engine Fuels 4.1 Category-A Jet Fuels Example Measurements Figure 4.1 shows a series of C 2H 4 mole fraction time-histories at different initial reflected shock temperatures. Clearly, the higher the temperature, the faster C 2H 4 grows. In addition, at the highest temperature, it appears that the C 2H 4 mole fraction may reach a plateau at long times. In later sections, the product yield is defined as molecules of product per equivalent molecule of fuel. C 2 H 4 Mole Fraction K 1260 K 1228 K 1200 K 1196 K 1168 K 1148 K 0.7% A2/Ar 12 atm Time ( s) Figure 4.1 A series of C 2H 4 measurements at different temperatures at 12 atm during pyrolysis of 0.7% A2 in argon. Figure 4.2 shows representative time-history data (C 2H 4 and CH 4) for one reflected shock wave experiment: pyrolysis of A2 fuel in argon at 1228 K and 12.4 atm. This type of experiment is the first of its kind in terms of simultaneous diagnostics of C 2H 4 and CH 4 during pyrolysis of heavy distillate fuels. Note that excellent signal-to-noise ratios (SNR) are achieved for C 2H 4, and good SNR for CH 4 is achieved, considering its weaker product of absorption cross section with number density. Uncertainty in the species concentrations is typically ±20% of signal. Evidently, the shape and magnitude of two profiles are dramatically different. C 2H 4 appears to rise rapidly in the early 500 µs and then to grow slowly, while CH 4 appears to increase at a slow and steady rate. In addition, the magnitude of CH 4 seems consistently lower than that of C 2H 4 59

86 by a factor of 3-4. It has been found that the global combustion behaviors of Category-A fuels is most sensitive to the C 2H 4 yield [69, 150], and hence primary attention should be paid to matching the C 2H 4 mole fraction time-histories. Mole Fraction % A2/Ar 1228 K 12.4 atm C 2 H 4 CH Time ( s) Figure 4.2 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during A2 pyrolysis. Smooth solid lines: simulations using the HyChem approach [150] (version 02, May 2016). Initial reflected shock conditions: 1228 K, 12.4 atm, 0.73% A2/argon. Simulations using the HyChem model are also included in Figure 4.2. CHEMKIN-PRO (Release 15141) [147] was, again, used to conduct these simulations, but with constant enthalpy and constant pressure thermodynamic assumptions (since the measured pressure trace, though not shown, appeared constant during the pyrolysis). Clearly, the simulated C 2H 4 or CH 4 profile using the HyChem model is in good agreement with the measured one. This is expected, as discussed earlier in Section 1.2.4, the HyChem model is constrained by experimental data, including the shock-tube species sensing experiments as shown in Figure 4.2. Since J11 surrogate model, mentioned in Table 3.6, was not tuned against current IDT or species sensing data, comparisons of measured and simulated J11 pyrolysis product yields can help us gauge the performance of the surrogate model in more details. For example, Figure 4.3 compares the measured C 2H 4 yield time-history with J11 surrogate modeling prediction in one J11 pyrolysis experiment. The J11 surrogate modeling underpredicts the measurement, even though it does well in reproducing the IDT of J11 as shown in Figure 3.23 and Figure It is worth mentioning that species sensing is not limited to only C 2H 4 and CH 4 diagnostics shown in current studies. Many reactive flow applications, such as some advanced engine concepts 60

87 with Exhaust Gas Recirculation (EGR), require reliable predictions of combustion products profiles [1], and species sensing is of fundamental value in the development and validation of the combustion kinetics model used in those predictions % J11/Ar 1231 K 14.6 atm measured C 2 H 4 Product Yield simulated Time ( s) Figure 4.3 Laser absorption measurements of C 2H 4 during J11 pyrolysis. Initial reflected shock conditions: 1231 K, 14.6 atm, 0.26% J11/argon. Smooth dashed lines: simulations using the J11 surrogate model (see Table 3.6) Product Yields A large series of measurements like the one shown in Figure 4.2 were conducted for a range of temperatures for the three Category-A fuels at around 12 atm. A summary of the product yields, including both C 2H 4 and CH 4, at three time milestones (i.e. 0.5 ms, 1.0 ms, and 1.5 ms) are shown in Figure

88 2 0.7% Cat-A /Ar 12 atm A1 A2 A3 C 2 H Yield at 0.5 ms 1 CH 4 18 Carbon in C 2 H 4 (%) 9 Carbon in CH 4 (%) Temperature (K) % Cat-A /Ar 12 atm A1 A2 A3 C 2 H Yield at 1.0 ms 1 CH 4 18 Carbon in C 2 H 4 (%) 9 Carbon in CH 4 (%) Temperature (K) % Cat-A /Ar 12 atm C 2 H 4 A1 A2 A Yield at 1.5 ms CH Carbon in C 2 H 4 (%) 9 Carbon in CH 4 (%) Temperature (K) 0 0 Figure 4.4 C 2H 4 (blue) and CH 4 (red) product yields and associated carbon fractions during Category-A fuel pyrolysis; yields at 0.5 ms in TOP panel, yields at 1.0 ms in MIDDLE panel, and yields at 1.5 ms in BOTTOM panel. Dots represent individual experiment. Solid lines: HyChem simulations. Average reflected shock pressure and fuel loading: 12.4 atm and 0.73% fuel in argon. Several observations can be made about these measurements. First, interestingly, the product distributions for C 2H 4 and CH 4 for all three fuels are very similar, despite the fact that the three Category-A fuels have different compositions and molecular weight (see Table 1.2). Recall that the high-temperature ignition delay times of these fuels are also similar (see Figure 3.2, Figure 3.14). This might indicate that the Category-A fuels share some common functional groups that 62

89 are responsible for governing global combustion kinetics behaviors, as suggested by Dryer et al. [54, 153]. Second, it appears that there is more significant carbon conversion from the parent fuel to C 2H 4. More specifically, as shown in Figure 4.4, at the highest temperature tested, approximately 36% of the fuel carbon is converted to C 2H 4, while only 8% converted to CH 4. It is found through other detailed chemical kinetics simulations [53] that the remainder of the fuel carbon typically flows into other small alkenes (e.g., propene, iso-butene), small alkanes (e.g., ethane) and aromatics (e.g., benzene, toluene). To recover more of the fuel carbon, more target-specific diagnostics are needed [140]. Comparison of product yields as a function of temperature between measurements and simulations is also done in Figure 4.4. The HyChem model predicts the C 2H 4 reasonably well at below 1250 K, but still shows some underprediction by around 20% at above 1250 K. Additionally, CH 4 yield simulations appear to follow the continued increasing behavior shown in the data, with the HyChem model only showing slight underprediction in the 1.0 ms case and 1.5 ms case. This suggests that, although current data are directly used for the HyChem model development, more refinement and iterations are needed to make the model even better Evolution of Data In the Category-A pyrolysis campaign in 2014, it was assumed that a 1-λ C 2H 4 diagnostic (i.e., µm only) was sufficient to measure C 2H 4 yield. However, this was later found not entirely accurate, especially for fuel loading greater than 0.2%. This might be related to nonnegligible interference absorption at higher fuel loadings. Therefore, the Category-A dataset has been corrected to account for the (unmeasured) interference absorption. New experiments in 2016 with both online and offline measurements were analyzed to provide guidance in the correction. For example, Figure 4.5 shows two example comparisons of online-only and offlinecorrected C 2H 4 yield time history. The offline-corrected time history appears consistently lower than the online-only one by approximately 20%, and hence a 20% correction, to the first order and for simplicity, was exerted on the Category-A dataset (all data in Figure are already correction-incorporated). Admittedly, this correction still is rough and lacks solid scientific evidence, and further work is underway to better resolve this issue. 63

90 C 2 H 4 yield online only on - off 0.75% A2/ Ar 1180 K 16.3 atm C 2 H 4 yield online only on - off 0.69% A2 / Ar 1370 K 15.4 atm Time ( s) Time ( s) Figure 4.5 Comparisons of online-only and offline-corrected C2H4 yield time history. Initial reflected shock condition: LEFT, 0.75%A2/Ar, 1180 K, 16.3 atm; RIGHT, 0.69% A2/Ar, 1370 K, 15.4 atm. 4.2 Category-C Jet Fuels Example Measurements The pyrolysis campaign of Category-C jet fuels took place in Spring 2016 using the 2-λ (differential) method for measurements of both C 2H 4 and CH 4. Figure 4.6 shows three example simultaneous C 2H 4 and CH 4 time-history measurements: pyrolysis of C6 fuel in argon at 1387 K and 12.3 atm, pyrolysis of C5 fuel in argon at 1274 K and 12.8 atm, and pyrolysis of C1 fuel in argon at 1375 K and 14.3 atm, respectively. Good signal-to-noise ratios are achieved for both species. Uncertainty in the species concentrations is estimated to be approximately ±25%, considering uncertainties propagated from both on-line and off-line signals. As seen in Figure 4.6 a and b, C 2H 4 appears to rise rapidly in the early 500 µs and then to increase more gradually, while CH 4 appears to grow at a slow and steady rate. This behavior is very similar to that shown in Figure 4.2. In addition, for both C6 and C5, the magnitude of CH 4 seems consistently lower than that of C 2H 4 by a factor of 3-4. This behavior is, again, similar to that of Category-A fuels. Figure 4.6 c, however, shows a dramatically different behavior of C 2H 4 and CH4 timehistories. C1, a fuel predominantly composed of iso-paraffins (25% C 16H 34 and 75% C 12H 26), yields only trace amounts of C 2H 4, and appears to produce slightly more CH 4 than C 2H 4, during its pyrolysis. This clearly demonstrates that for some non-petroleum-derived fuels of unique molecule structure, C 2H 4 remains no longer the dominant pyrolysis product (one might argue that in the case of C1, C 2H 4 diagnostic loses the power of capturing the most valuable 64

91 information during fuel decomposition; however, it is this low-c 2H 4-output feature that has inspired our group to develop and implement multi-species diagnostics, such as propene and isobutene, for further investigation of pyrolysis of C1, see ref [140]) % C6/Ar 1387 K 12.3 atm C 2 H % C5/Ar 1274 K 12.8 atm % C1/Ar 1375 K 14.3 atm Mole Fraction CH 4 Mole Fraction C 2 H 4 CH 4 Mole Fraction CH 4 C 2 H Time ( s) Time ( s) Time ( s) a b c Figure 4.6 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during C6 (a), C5 (b), and C1 (c) pyrolysis. Simulations for C5 and C1 using the version 2 of the HyChem model are shown in smooth solid lines. Initial reflected shock conditions: 1387 K, 12.3 atm, 0.6% C6/Ar; 1274 K, 12.8 atm, 0.65% C5/Ar; 1375 K, 14.3 atm, 0.71% C5/Ar Product Yields Measurements like the ones shown in Figure 4.6 were systematically conducted for all six Category-C fuels at around 12 atm and a range of temperatures. A summary plot of the product yields, including both C 2H 4 and CH 4, at an early time (i.e. 0.5 ms) is shown in Figure 4.7. C 2 H 4 Yield at 0.5 ms C1 C2 C3 C4 C5 C Temperature (K) 65

92 CH 4 Yield at 0.5 ms C1 C2 C3 C4 C5 C Temperature (K) Figure 4.7 C 2H 4 (shown in the top panel in squares) and CH 4 (shown in the bottom panel in circles) product yields during Category-C fuel pyrolysis; yields at 500 µs. Average reflected shock pressure and fuel loading: 12 atm and 0.6% fuel in argon. The solid lines are simply smoothed fits to the data. It is interesting to see from these measurements how the product yields vary from fuel to fuel. First, it appears that C2, C3, C5, and C6 have similar level of C 2H 4 yields, C4 has less by around 50%, and C1 shows the lowest C 2H 4 yields. This might be explained by comparing the hydrocarbon class composition of these fuels. C1 is 100% Gevo Alcohol-To-Jet (ATJ) fuel whose constituents are 99% C 12 and C 16 iso-paraffins, and kinetic simulations show that isobutene, instead of ethylene, is the dominant decomposition product of iso-paraffins [67]. Similarly, C4 also exhibits low level of C 2H 4 yields, but slightly more than those of C1, possibly due to the lower amount of Gevo ATJ (40%). The remaining Category-C fuels are not as short of C 2H 4-producing components as C1 and C4. Another observation from Figure 4.7 is that while most fuels appear to show monotonic increase of C 2H 4 yields with increasing temperature, C2 appears to exhibit a slight roll-over phenomenon at the highest temperature end, i.e K. This roll-over might be due to conversion of C 2H 4 to C 2H 2, and might suggest that this conversion occurs at a lower temperature for C2 as compared to other Category-C fuels, but more data are needed to confirm this apparent trend. Second, all six fuels show similar magnitude (within 50%) and shape of CH 4 yields for the current temperature range. Note that there is evidently continued formation at higher temperatures (except C1). This is likely to happen as other products begin to slowly decompose 66

93 to CH 4, as a result of the long life-time of CH 3 at high temperatures and the balance of C-H ratio among the decomposition products. Lastly, current species diagnostics capture different levels of total carbon conversion (30% - 60%) for different fuels. More specifically, at the highest temperature tested, the percentages of fuel carbon converted to C 2H 4 and CH 4 for the six fuels, in descending order of the total carbon captured, are shown in Table 4.1. It is found that the remainder of the fuel carbon typically flows into other alkenes and aromatics, as discussed above for Category-A fuels. To recover more of the fuel carbon, especially for C1 and C4, more target-specific diagnostics are needed, as shown in a related work in our laboratory [140]. Table 4.1 Carbon fractions captured in C 2H 4 and CH 4 diagnostics during pyrolysis of Category-C fuels at near 1600 K and 0.5 ms. Fuel Carbon in C2H4 Carbon in CH4 C5 52% 13% C3 47% 16% C6 42% 13% C2 39% 6% C4 22% 11% C1 19% 10% Figure 4.8 shows a comparison of measured and simulated C 2H 4 and CH 4 yields of the selected Category-C fuels, C1 and C5. Simulations use the Version 2.0 of the HyChem model. Given the uncertainties of current C 2H 4 and CH 4 speciation, it is reasonable to conclude that the simulations acceptably capture both the shape and the magnitude of these data. 67

94 Yield at 0.5 ms C 2 H 4 CH 4 Fuel Data: Modeling: C1 C5 C1 C Temperature (K) Figure 4.8 C 2H 4 and CH 4 product yields during C1 and C5 fuel pyrolysis; yields at 500 µs. Data are shown in squares to C 2H 4 and circles for CH 4, respectively. Modeling was done using the version 2 of the HyChem mechanism with C 2H 4 yield modeling shown in solid lines and CH 4 yield modeling shown in dashed lines. Average reflected shock pressure and fuel loading: 12 atm and 0.6% fuel in argon. 4.3 RP-2 Fuels Example Measurements The pyrolysis campaign of two RP-2 fuels (R4 and R5) took place in Spring 2016 using the 2- λ (differential) method for measurements of both C 2H 4 and CH 4. Figure 4.9 compares the simultaneously measured C 2H 4 and CH 4 mole fraction time-histories with the HyChem simulations. Note, again, the good SNR in these signals and the good agreement between data and modeling. The uncertainty of current RP-2 pyrolysis measurements is around 30%. 68

95 Mole Fraction % R4/Ar 1387 K 12.8 atm C 2 H 4 CH Time ( s) Figure 4.9 Laser absorption measurements of C 2H 4 (in blue) and CH 4 (in red) during R4 pyrolysis. Simulations using the version 2 of the HyChem model are shown in smooth solid lines. Initial reflected shock conditions: 1387 K, 12.8 atm, 0.68% R4/Ar Product Yields The same methodology for studying jet fuels was used for studying RP-2 fuels, with the C 2H 4 and CH 4 yields data at different temperatures shown in Figure The HyChem simulations are also included. Interestingly, the data appear to show that the C 2H 4 yield or CH 4 yield of R5 is generally higher than that of R4, with the C 2H 4 yield exhibiting the largest difference by approximately 30%. In addition, the HyChem modeling appears to also capture the differences in the C 2H 4 yield of R4 and R5. For modeling of C 2H 4 yield, the R5 simulations appear to show better agreement with data than the R4 simulations do. For modeling of CH 4 yield, both R4 and R5 modeling appear to well reproduce the experiments. 69

96 4 C2 H 4 CH 4 Fuel Yield at 0.5 ms 2 Data: R4 R5 Modeling: R4 R Temperature (K) 4 C2 H 4 CH 4 Fuel Yield at 1.0 ms 2 Data: R4 R5 Modeling: R4 R Temperature (K) Yield at 1.5 ms 4 C2 H 4 CH 4 Fuel 2 Data: R4 R5 Modeling: R4 R Temperature (K) Figure 4.10 C 2H 4 and CH 4 product yields during R4 (shown in black) and R5 (shown in red) pyrolysis. Dots represent individual experiment with solid squares for C 2H 4 and hollow circles for CH 4. Solid lines: HyChem (version 2) simulations for C 2H 4 yields. Dashed lines: HyChem (version 2) simulations for CH 4 yields. Average reflected shock pressure and fuel loading: 13.5 atm and 0.76% fuel in argon. 70

97 5 Chapter 5 Kinetic Studies of Decalin 5.1 Ignition Delay Time Figure 5.1 and Figure 5.2 show representative ignition delay time measurements at high and low temperatures, respectively, and corresponding simulations using the semi-detailed chemical kinetic mechanism for decalin oxidation published by Dagaut et al. [82] using the conventional constant-volume (V), constant-energy (U) assumption. All simulations were performed with the OpenSMOKE code [154, 155]. Figure 5.1 shows representative pressure traces for both a nonreactive (pure N 2) case and reactive (decalin/o 2/N 2) case and the OH* emission record. In the current study, ignition delay time is defined as the time interval between the arrival of reflected shock at the measurement location and the onset of ignition determined by extrapolating the maximum slope of either pressure or OH* record back to the baseline. As seen in Figure 5.1, the pressure trace and OH* record exhibit consistent ignition delay times within ±1%. In addition, the pressure trace from simulation of this experiment is plotted for comparison. Both measured and simulated pressure traces exhibit rapid and smooth exponential rises upon ignition, although slight pre-ignition pressure increase is evident in the simulation compared to the pressure trace in the non-reactive case. Pressure (atm) Pressure (reactive) Pressure (non-reactive) Pressure (const. UV simulation) 306 nm OH* emission 1.62% Decalin/20.7% O 2 /N K, 22.1 atm, = 1.1 t ign emission (a.u.) t ign Time ( s) Figure 5.1 Example ignition delay time measurement at high temperature. The reaction mechanism used for simulating this measurement was published in [82]. 71

98 Representative pressure traces for lower temperatures are presented in Figure 5.2. The pressure trace in the non-reactive experiment confirms that no reflected-pressure gradient (dp 5/dt) is evident in the long-test-time ignition experiments. However, significant pre-ignition pressure rises are clearly evident in the reactive experiments. These rises manifest themselves as either a ramp till the instant of ignition (830 K case) or a step before the final ignition event (802 K case). It is important to report these pressure traces associated with pronounced pre-ignition heat release when simulating shock tube experiments containing non-dilute mixtures, as the traditional constant U, V assumption might not be valid in the presence of pre-ignition pressure increases [156]. The goodness of this assumption can be evaluated by comparing the measured and simulated pressure profiles. Fortunately, simulations show very good agreement with the experiments, both qualitatively and quantitatively. Not only do the simulated pressure traces reproduce the behavior observed in the experiments (ramp or step), but also the predicted ignition delay times are very close to the measured ones based on the definition of ignition delay time mentioned above. Such agreement implies that the pre-ignition pressure rises shown in current measurements result from pre-ignition chemistry and justifies the constant U, V assumption. Nonetheless, to circumvent the controversy associated with the pre-ignition pressure rises, the innovative constrained-reaction-volume strategy [128, 156] will be employed in future work to revisit the autoignition of decalin/air mixtures, but at near-constant-pressure conditions Reactive Non-reactive Const. UV simulations Decalin/air Pressure (atm) t ign 830 K 24.0 atm = K 19.3 atm = t ign Time (ms) Figure 5.2 Example ignition delay time measurements at low temperature. The reaction mechanism used for simulating this measurement was published in [82]. 72

99 All ignition delay data are classified into high-temperature ( K) and low-temperature ( K) regimes and summarized in the appendix (only those determined from pressure traces are reported). The uncertainties in the reflected shock temperature and pressure and initial decalin concentration are estimated to be 0.7%, 1.5%, and 5%, respectively, so the uncertainty of the current ignition delay time data is estimated to be 15% by the theory of propagation of uncertainty. An ignition delay time correlation will be derived based on only high-temperature data and used to scale the data in the high-temperature regime. The low-temperature data will be scaled by a temperature-dependent correlation derived from simulations to be discussed later High-temperature Ignition Figure 5.3 shows ignition delay data of a stoichiometric decalin/air mixture at various pressures from two sources of experiments. RPI refers to data measured by Oehlschlaeger et al. [96] and SU stands for data in the current study. Both sources of data are scaled to nominal pressures and equivalence ratios using individually reported correlations. The pressure-scaling dimensions are in excellent agreement, i.e., t ign - P -0.78, while the global activation energy from RPI s data (124 kj/mol) is slightly higher than that from the current study (110 kj/mol). There is some evidence in the 12 atm data at low temperatures that the Stanford ignition delay times are slightly shorter than those from RPI = %Decalin/21% O 2 /N 2 T (K) RPI 12atm 1000 SU 12atm SU 20atm RPI 40 atm 909 SU 50atm t ign ( s) /T (1/K) Figure 5.3 Comparison of ignition delay data of stoichiometric decalin/air mixture at various pressures from RPI [96] and SU (current studies). Solid and dashed lines are best fits to data points. 73

100 Figure 5.4 and Figure 5.5 present variation of ignition delay time of decalin in air with equivalence ratio. As is evident in the two sources of data, the equivalence ratio dependence is negative over the pressure range of 12 to 50 atm. However, the equivalence ratio scaling based on RPI s data is t ign - ϕ -0.81, while the current data support t ign - ϕ Given the negative dependence, if fuel loss were to occur, the actual equivalence ratio of the mixture in the ignition event would be lower than expected, and hence the ignition delay time would be longer. This might partially explain why current data appear shorter than those of RPI at the same condition. Specifically, due to the low saturated vapor pressure of decalin, it is plausible that the fuel concentration in the shock tube might be lower than that prepared manometrically in the mixing tank. In other words, a portion of the fuel may be lost somewhere along the fuel line, either by condensation or adsorption. Since the initial fuel concentration was not monitored in RPI s studies, it is unsure whether they had such fuel-loss problem. In our current studies with this fuel, typical fractional fuel loss (i.e., the difference between the manometrically determined fuel concentration in the mixing tank and the in situ fuel concentration in the shock tube measured using laser absorption) was %. Similar losses may be expected in other shock tube facilities T (K) RPI, 12 atm, = 0.5 = 1.0 SU, 20 atm, = 0.5 = 1.0 = 1.5 t ign ( s) /T (1/K) Figure 5.4 Comparison of ignition delay data of decalin/air mixture at various equivalence ratios from RPI [96] at 12 atm and SU (current studies) at 20 atm. Solid and dashed lines are best fits to data points. 74

101 T (K) SU, 50 atm, = = 1.0 RPI, 40 atm, = 0.5 t ign ( s) 100 = /T (1/K) Figure 5.5 Comparison of ignition delay data of decalin/air mixture at various equivalence ratios from RPI [96] at 40 atm and SU (current studies) at 50 atm. Solid and dashed lines are best fits to data points. Our measured ignition delay times at these high temperatures and wide ranges of pressure and equivalence ratio can be correlated through least-squares regression as t ign (µs) = P ϕ exp(110 kj/mol/rt) Eqn. 5 where t ign is the ignition delay time in micro seconds, T the temperature in Kelvin, and P the pressure in atm. This correlation is plotted in Figure Decalin/Air t ign ( s) /T (1/K) Figure 5.6 Ignition delay time data for decalin/air at above 1000 K. All data are normalized to 20 atm and ϕ = 1.0 using the scalings of Eqn. 5. The Eqn. 5 correlation is shown as a solid line. 75

102 Figure 5.7 and Figure 5.8 compare current ignition delay times at high temperatures with kinetic modeling using the aforementioned mechanism [82] (Ranzi modeling for short). The Ranzi modeling generally overpredicts the ignition delay time of decalin at high temperatures by at most a factor of 2. In addition, while these mechanistic predictions yield a similar global activation energy (107 kj/mol) to the measured value (110 kj/mol), they show slightly stronger dependence on pressure (t ign - P ) and weaker dependence on equivalence ratio (t ign - ϕ ) compared to current findings (t ign - P ϕ ). Due to the propagation of low-temperature chemistry into high-temperature regimes at elevated pressures [157], it is possible to achieve better agreement between experiments and modeling in Figure 5.7 and Figure 5.8, as will be discussed in a later section T (K) Dashed lines: Ranzi modeling Dots: SU data Solid lines: best fits to data 12 atm 20 atm atm t ign ( s) 100 Decalin/21% O 2 /N 2 = 1.0, X f = 1.4% /T (1/K) Figure 5.7 Comparison of measured and originally predicted ignition delay data of stoichiometric decalin/air mixture at 12, 20, and 50 atm over K. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. 76

103 T (K) Dashed lines: Ranzi modeling Dots: SU data Solid lines: best fits to data 1000 t ign ( s) 100 Decalin/21% O 2 /N 2 P = 20 atm /T (1/K) Figure 5.8 Comparison of measured and originally predicted ignition delay data of decalin/air mixture at ϕ = 0.5, 1.0, and 2.0, 20atm over K. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption Low-temperature Ignition Figure 5.9 exhibits the negative-temperature-coefficient (NTC) behavior of decalin and its dependences on pressure and equivalence ratio. This NTC behavior is a well-acknowledged but incompletely characterized aspect of straight-chain alkanes [80, 81], branched-chain alkanes [78, 79], and some jet fuels [8, 44]. Similar NTC behavior for decalin oxidation is also noted by Agosta [36] in terms of carbon monoxide production. As shown in Figure 5.9, the widths of the NTC regime at three different conditions are quite close and relatively narrow. In addition, the start of the NTC regime at ϕ = 1.0 and 50 atm occurs at higher temperature (around 940 K) compared to that at ϕ = 1.0 and 20 atm (around 880 K), while the starting point at ϕ = 0.5 and 1.0, 20 atm are nearly the same. This pressure-induced shift is a typical aspect of NTC behavior and has been related to the equilibria of the addition reactions of oxygen to the alkyl (R) and hydroperoxy-alkyl (QOOH) radicals [158]. Furthermore, the influence of pressure and equivalence ratio on the ignition delay time of decalin/air mixture is most pronounced in the NTC regime, but cannot be quantitatively described by simple power law correlations as discovered at higher temperatures. The pressure and equivalence ratio dependences of the NTC regime for decalin autoignition are qualitatively in good agreement with those of n-heptane [113] and n-decane [114]. 77

104 Kinetic model predictions using decalin oxidation mechanism [82] and the constant U, V assumption are also presented in Figure 5.9. Comparisons between measurements and modeling illustrate that Ranzi modeling captures many of the experimental trends. First, the overall agreement between data and modeling is surprisingly good, with better agreement at lower temperatures for data at all three conditions. Second, the Ranzi-predicted pressure and equivalence ratio dependences also appear to be strongest in the NTC regime. These predicted dependences are employed to scale the experimental data at K by the expression t ign - P m ϕ n where m and n are functions of temperature, as shown in Figure Nonetheless, some deviations in the experiments and modeling still exist. For example, the observed global activation energy in the NTC regime at ϕ = 1.0 and 50 atm appears to be slightly negative, while the predicted one is positive. Although these differences are relatively small, improvement to modeling predictions is still possible but will require examination of the reactions responsible for both high- and low-temperature autoignition T (K) Decalin/21% O 2 /N 2 Dashed lines: Ranzi modeling Dots: SU data Solid lines: best fits to data t ign (ms) 1 =0.5, 20 atm =1, 50 atm =1, 20 atm /T (1/K) Figure 5.9 Comparison of measured and originally predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 20 atm; ϕ = 1.0, 20 atm; ϕ = 1.0, 50 atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. NTC behavior is evident at around K. 78

105 K 833 K 769 K n P m Decalin/air Ranzi modeling t ign ~ P m n m or n /T (1/K) Figure 5.10 Temperature-dependent pressure and equivalence ratio scaling dimensions for ignition delay times at K obtained from Ranzi modeling. 5.2 Pyrolysis Fast fuel cracking at early times allows decoupling of the fuel decomposition chemistry from the oxidation kinetics of the cracked products [159]; thus, it is feasible to focus in separate experiments on the chemistry of decalin pyrolysis. Pyrolysis experiments were carried out using a mixture of decalin/argon at atm over a temperature range of K within test times of -2 ms. Before presenting the experimental results, it is useful to discuss the major products of decalin pyrolysis at these experimental conditions. Figure 5.11 shows an example kinetic modeling of decalin pyrolysis at a representative shock tube condition using the same reaction mechanism and gasdynamic assumption as in the autoignition cases. It reveals that the expected primary products are ethylene, benzene, propene, toluene, and 1,3-butadiene (the order is contingent on the specific experimental conditions). Also note that one intermediate species, ODECAL, using the notation of Ranzi which represents a group of cyclo-alkene species, indicates the lumping nature of the reaction scheme. Additionally, besides ethylene, it is not surprising to have benzene and toluene in the list of most prevalently formed products. In fact, the decalin pyrolysis system has received much attention because it is a significant feedstock hydrocarbon processed by the petroleum industry to produce other hydrocarbons, especially more aromatics than from pyrolysis of alkanes or mono-cyclic five- and/or six -membered cyclo-alkanes [25]. This list of primary products is confirmed by the flow reactor studies by Zeppieri et al. [91] and the ab initio study by Chae et al. [99] at similar conditions. 79

106 In the current study, the 3.39 μm diagnostic provides an overall check of the performance of the model for decalin pyrolysis, and the 10.6 μm diagnostic is conducive to a better understanding of products distribution, at least partially through ethylene mole fraction time-histories. The initial decalin concentrations utilized in the 3.39 μm and 10.6 μm experiments were around 2260 ppm and 3540 ppm, respectively C 2 H 4 Mole Fraction (ppm) C 6 H 6 C 3 H 6 C 7 H 8 C 4 H 6 C 5 H 8 DECALIN Ranzi modeling 2260 ppm Decalin/Ar ODECAL 1350 K, 19.0 atm Time ( s) Figure 5.11 Example kinetic modeling of decalin pyrolysis at a representative shock tube condition using the reaction mechanism in [82] and constant U, V assumption μm Diagnostic Figure 5.12 shows time-histories of 3.39 μm absorbance from five decalin pyrolysis experiments. Time zero is defined as the moment of arrival of the reflected shock at the measurement location. The indicated temperatures refer to the calculated values immediately behind the reflected shock. Although these absorbance time-histories generally exhibit exponential-decay-like behaviors, caution must be exercised that they represent the total absorbance from all hydrocarbons involved in the process of decalin pyrolysis, much more than those depicted in Figure Therefore, it is not easy to extract information with respect to the behavior of the parent fuel, i.e., decalin. Instead of correcting the measured absorbance by either a simple one-step-reaction method [75] or a detailed kinetic mechanism method [51], we propose a simple approach to utilizing these measurements to evaluate the decalin pyrolysis kinetics. With a knowledge of absorption cross section database [51, 76, 131, ] of hydrocarbons at high temperatures and pressures at a particular wavelength, we can convert the predicted time-histories of species mole fraction to those of absorbance and compare the sum of these predicted absorbances with 80

107 the measured ones. A summary of the database used in current studies is shown in Table 5.1. Since this approach requires a good prediction of the products distribution, the 10.6 μm diagnostic was employed to help impose some constraints on the decalin pyrolysis submechanism m Absorbance K 1233 K 1297 K 1369 K 2260 ppm Decalin/Ar atm K Time ( s) Figure 5.12 Absorbance time-histories at 3.39 μm during pyrolysis of 2260 ppm decalin/ar at K and atm. Table 5.1 Absorption cross sections of major hydrocarbons at 3.39 and μm in the decalin pyrolysis system. Fits listed here were derived from cross section measurements at high T and P. Absorption Cross Section (m 2 /mol) Species 3.39 μm μm References DECALIN *T E-6*T this work;[161] a ODECAL *T-9.41E-5*T E-8*T 3 4 [76] b ;[76] c C 2H E-4*T 4 [51];[76] C 6H [161];[161] C 3H *T *T [51];[76] C 7H *T 0.36 [162];[162] C 4H [161];[161] C 5H [161];[161] CH [160], [131];[161] a taken as cyclohexane b taken as methylcyclohexane c taken as ethylene 81

108 μm Diagnostic The two-color differential absorbance method developed by MacDonald et al. [76] was employed to extract the time-history of ethylene mole fraction. This method requires separate experiments, usually first at μm and then μm, such that the interference absorption can be cancelled out. To have a sense of carbon conversion from decalin to ethylene, the ethylene mole fraction is normalized by the initial mole fraction of decalin to give rise to the ethylene yield. The uncertainty of current ethylene measurements comes from several sources including the original absorbance signals, reflected-shock temperature and pressure, and absorption cross sections of ethylene at two colors, with the first one being dominant due to relatively larger beam steering noise of the CO 2 laser beam. The combined uncertainty of the experimental ethylene yield is estimated to be 20-30%. Figure 5.13 presents measured and predicted ethylene yield traces for four decalin pyrolysis experiments. We find that the data and modeling are in good agreement only in terms of shape but not magnitude. Specifically, the Ranzi modeling overpredicts both the early time rise and long-time plateau of the ethylene yield. This might imply that although the ethylene formation pathways are reasonably conceived, the corresponding rate parameters need refinement Solid: Current measurements Dashed: Ranzi modeling ~3540 ppm Decalin/Ar atm C 2 H 4 Yield K 1412K 1309K 1218K Time ( s) Figure 5.13 Comparison of measured and originally predicted ethylene yield time-histories during pyrolysis of ppm decalin/ar at K and atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. 82

109 Comparison of the long-time plateau of ethylene yield, or peak ethylene yield, is helpful to assess the ability of decalin to produce ethylene. Figure 5.14 shows the observed and predicted peak ethylene yields as a function of temperature. The observed values at 1218 K and 1309 K are obtained through exponential fit to the corresponding ethylene yield for the first 2 ms. The Ranzi modeling generally overpredicts the peak ethylene yield at these temperatures by around 50%. In addition, current measurements indicate that ethylene accounts for only 20 25% of the total carbon in the parent decalin molecule, which means that the yields of some interfering species, most probably other small alkenes (propene and 1,3-butadiene) or aromatics (benzene and toluene), are comparable to, or even higher than that of ethylene. Since these interfering species make significant contributions to the absorbance measured at μm [76], adjustment of the decalin pyrolysis sub-mechanism should enable us to recover both the experimental ethylene yield profiles and the measured absorbance at μm. Peak Ethylene Yield Current studies Ranzi modeling Upper-limit estimation ~3540 ppm Decalin/Ar atm T (K) Carbon Conversion (%) Figure 5.14 Comparison of measured, predicted and estimated peak ethylene yield during pyrolysis of ppm decalin/ar at K and atm. Mechanism predictions, or Ranzi modeling, were calculated using the reaction mechanism [82] and constant U, V assumption. Upper-limit estimation were derived based on analysis of the reaction pathways in Figure 5.15 (see text). 5.3 Understanding the Decalin Kinetics The goal of current measurements reported above is to validate and refine the decalin reaction mechanism. Since the Ranzi decalin model uses a lumped reaction/species approach, it might not be possible to make direct and specific kinetic adjustment based on a detailed analysis of 83

110 individual pathways, species sensitivities, and reaction fluxes, etc.. Nonetheless, some preliminary suggestions can still be made within this framework of lumping reaction scheme, as will be discussed below Pyrolysis Kinetics Before exploring possible improvement to the mechanism, it is necessary to check with other decalin pyrolysis mechanisms to confirm the overprediction aspect of Ranzi modeling. To the best of our knowledge, there exist only two other mechanisms that describe the pyrolysis of decalin at relatively high temperatures (above 1000 K), i.e., the computational kinetic mechanism developed based on and validated against the flow reactor studies by Zeppieri [25] and the ab initio study of thermal decomposition of decalin [99], mentioned previously. Unfortunately, we have found that the former does not yield reasonable results at the current shock tube experimental conditions. This is not surprising as parameters important in an intermediate-temperature flow reactor experiment might not be the same as those in a hightemperature shock tube simulation [163]. In addition, the reaction scheme and thermochemical data assembled in [99] are not entirely available, so a direct use of this model cannot be made. Nonetheless, an analysis of the reaction scheme released in [99] inspires us to estimate the upper-limit of ethylene formation at current experimental conditions. Specifically, consider the initiation kinetics of decalin pyrolysis shown by pathways below (see Figure 5.15). Pathway C- C refers to the breaking of the bridgehead bond of decalin molecule, and pathways 1-1, 10-1, and 12-1 represent the hydrogen-abstraction reactions at three types of carbon sites in the decalin molecule and subsequent reactions leading to the formation of ethylene, benzene, etc.. If we only take into account pathways 1-1 and 12-1, the upper-limit of ethylene formation can be deduced. Although pathway C-C can also give birth to ethylene [99], it has been experimentally studied [89-91] and mechanistically argued [25] that the bridgehead bond homolysis route is a minor route of decalin decay. This route is not included in the decalin sub-mechanism published in [82], either. Therefore, pathway C-C is neglected in current endeavor of upper-limit estimation. Results of this estimation are also shown in Figure Clearly, the estimated upper-limits of peak ethylene yield are nearly identical with those obtained from Ranzi modeling (within 6%). This justifies the necessity of adjusting the decalin pyrolysis submechanism released in [82]. 84

111 Figure 5.15 Initial decomposition pathways of decalin adapted from [99] (using the same reaction labels). Only pathways 1-1 and 12-1 were included in the estimation of upper-limit of peak ethylene yield. To seek guidance, A-factor sensitivity analysis for ethylene was conducted, as shown in Figure Generally, ethylene formation is most sensitive to several reactions that compete with each other. At very early times (~100 μs), it is very sensitive to the isomerization reaction of decalin to ODECAL, a group of cyclo-alkene species as mentioned before, while at long time, its fate is mostly determined by the two competing decomposition reactions of ODECAL, i.e., reactions pyr 1 and pyr 2 (shown in Figure 5.16) with rate constants k pyr 1 and k pyr 2, respectively. Although k pyr 1 and k pyr 2 have been adjusted to explain the large formation of benzene and toluene [82], the actual branching ratios and reaction rate constants of these initial individual channels and the subsequent reactions that lead from these intermediates to stable final products are still somewhat a matter of conjecture. Hence, it is still possible to make improvement, at least to reproduce the experimental ethylene yield reported here. C 2 H 4 A-factor Sensitivity Ranzi Modeling 3493 ppm Decalin/Ar 1412 K, 18.8 atm ODECAL=>2C2H4+C6H6+2H2 pyr 1 DECALIN=>CYC6H10+2C2H4 DECALIN=>ODECAL pyr 3 DECALIN=>2NC5H9-4 ODECAL=>C3H6+C7H8+2H2 pyr Time ( s) Figure 5.16 A-factor sensitivity analysis for ethylene during pyrolysis of 3493 ppm decalin in Ar at 1412 K and 18.8 atm. 85

112 As discussed earlier, adjustment of k pyr 1 and k pyr 2 should make the modeling match both the ethylene yield and μm absorbance. Two options are available: the first one is to multiply k pyr 1 and k pyr 2 by the same factor maintaining the branching ratio k pyr 1/( k pyr 1+ k pyr 2), while the second one is to increase the branching ratio holding the sum k pyr 1+ k pyr 2 constant. The second option is chosen here, as current ethylene measurements are independent of the overall decomposition rate of ODECAL. After changing the branching ratio k pyr 1/( k pyr 1+ k pyr 2) from 76.9% to 35.0%, significant improvement of agreement is achieved between experimental results and modeling. This improvement is evident not only in terms of ethylene yield timehistories, but also of peak ethylene yields, as shown in Figure 5.17 and Figure 5.18, respectively, within the uncertainty limits of current measurements. Nonetheless, Figure 5.17 shows that the modified mechanism predictions still overpredict the early time rise of ethylene, which might be due to deficiency in the untouched reaction pyr 3, DECALIN ODECAL. C 2 H 4 Yield Solid: Current measurements Dashed: Modified Ranzi modeling 1511 K 1412 K 1309 K ~3540 ppm Decalin/Ar atm 1218 K Time ( s) Figure 5.17 Comparison of measured and newly predicted ethylene yield time-histories during pyrolysis of ~3540 ppm decalin/ar at K and atm. Modified Ranzi modeling was made using the reaction mechanism [18] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. 86

113 Peak Ethylene Yield Current studies Modified Ranzi modeling ~3540 ppm Decalin/Ar atm T (K) Carbon Conversion (%) Figure 5.18 Comparison of measured and predicted peak ethylene yield during pyrolysis of ~3540 ppm decalin/ar at K and atm. Modified Ranzi modeling was made using the reaction mechanism [18] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. Additionally, a comparison of peak ethylene yields of different classes of alkanes [76] is shown in Figure Clearly, the peak ethylene yields of n-dodecane (straight-chain alkane) are typically much higher than those of decalin and methylcyclohexane (cyclo-alkanes) by around a factor of four, and the latter two species have nearly the same peak ethylene yields, while those of iso-cetane (branched-chain alkane) are almost negligible. Such information is quite useful for formulation of chemical kinetic surrogates for practical fuels containing a large fraction of cyclo-alkanes [52]. Peak Ethylene Yield n-dodecane (MacDonald et al.) methylcyclohexane (MacDonald et al.) iso-cetane (MacDonald et al.) Decalin (this study) 1100~3540 ppm Fuel/Ar atm T (K) Figure 5.19 Comparison of peak ethylene yields of different classes of alkanes: n-dodecane [76], methylcyclohexane [76], iso-cetane [76], and decalin (this study). 87

114 Besides ethylene yield, the modified modeling is validated against the measured absorbance at μm in four decalin pyrolysis experiments, as shown in Figure Fortunately, the agreement is generally satisfactory, within 20% for the three lower temperature cases. At the highest temperature case, although the calculated absorbance overpredicts the long-time behavior, it still captures the peak of the measured absorbance. It is also noted that the primary interfering species to 10.6 μm diagnostic for decalin are propene and 1, 3-butadiene, which illustrates the promise of applying more infrared diagnostics under development in our laboratory to measure these interfering species for a pyrolysis system like that of decalin. Figure 5.20 Comparison of calculated (black) and measured (red) absorbances at μm during pyrolysis of ~3540 ppm decalin/ar at K and atm. Calculations were made using the reaction mechanism [82] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. Figures at right show zoomed-in view of those at left. Similar validation work can be performed with regard to the measured absorbance at 3.39 μm shown in Figure As mentioned before, the vibrational mode of stretch of C-H bond in any hydrocarbon is strongly resonant with the 3.39 μm laser light, so more species should be 88

115 considered in the validation at 3.39 μm, as shown in Table 5.1. Comparison of calculated and observed absorbances at 3.39 μm is shown in Figure The overall agreement is satisfactory (within 25%), given the limited set of species involved in the calculations and the fidelity of absorption cross section of ODECAL at high temperatures and pressures. It is learned that the primary contributor other than decalin to the 3.39 μm absorbance at the early 500 μs is ODECAL, while those at intermediate-to-long times are propene and ethylene. The contribution of methane to the total calculated absorbance increases with increasing temperature, while those from the remaining species considered are usually less than Such validation, in a global sense, tests the overall goodness of the modified decalin pyrolysis sub-mechanism. Further refinement could be made with the help of a diagnostic sensitive only to decalin as well as diagnostics for additional decomposition products. 89

116 Figure 5.21 Comparison of calculated (black) and measured (red) absorbances at 3.39 μm during pyrolysis of ~2260 ppm decalin/ar at K and atm. Calculations were made using the reaction mechanism [82] with modified rate constants for reactions pyr 1 and pyr 2 in Figure 5.16 and constant U, V assumption. Figures at right show zoomed-in view of those at left Oxidation Kinetics Figure 5.22 shows a brute-force sensitivity analysis of ignition delay time of decalin in air at 20 atm (top) and 50 atm (bottom), ϕ = 1.0. Sensitivities of ten reactions at 1250, 1000, and 833 K are examined. At the highest temperature at 20 atm, reaction oxi 1 plays the predominant role in the oxidation of decalin, as expected, while reactions oxi 3 and oxi 4 come second and third. At the lowest temperature at 20 atm, reactions oxi 6 and oxi 7 have dramatic influence on the ignition delay time. These reactions are typical examples of low-temperature autoignition chemistry [164], representing classes of R + O 2 RO 2 and QOOH + O 2 OOQOOH, respectively. Abstraction by hydroperoxy radical (HO 2), hydroxyl radical (OH), and oxygen (O 2) from the decalin molecule (reactions oxi 3, oxi 2, oxi 4, and oxi 5 ) also affects the final ignition delay time at 20 atm, in the decreasing order, and reactions oxi 2 and oxi 5 appear to be significant only at the lowest temperature at 20 atm. The three most influential reactions in the decalin pyrolysis system, i.e., reactions pyr 1, pyr 2, and pyr 3, have negligible impact on the ignition delay time at all three temperatures at 20 atm, which further justifies the decoupling of adjustments to the pyrolysis kinetics from those of the oxidation kinetics. Sensitivity of ignition delay time of decalin at 50 atm exhibits similar patterns as those at 20 atm, except the increasing importance of reactions oxi 6 and oxi 7 at 1000 K, which confirms the shift of NTC regime to higher temperature with increasing pressure observed in Figure

117 1250 K 1000 K 833 K Decalin/air 20 atm = 1 H + O2 <-> OH + O oxi 1 OH + DECALIN -> H2O + RDECAL oxi 2 HO2 + DECALIN -> H2O2 + RDECAL oxi 3 DECALIN -> ODECAL pyr 3 ODECAL -> 2C2H4 + Benzene + 2H2 pyr 1 ODECAL -> C3H6 + Toluene + 2H2 pyr 2 O2 + DECALIN -> HO2 + RDECAL oxi 4 O2 + DECALIN -> HO Tetralin DECALIN oxi 5 RDECAL + O2 -> RDECOO oxi 6 QDECOOH + O2 -> ZDECA oxi Sensitivity of ignition delay time 1250 K 1000 K 833 K Decalin/air 50 atm = 1 H + O2 <-> OH + O oxi 1 OH + DECALIN -> H2O + RDECAL oxi 2 HO2 + DECALIN -> H2O2 + RDECAL oxi 3 DECALIN -> ODECAL pyr 3 ODECAL -> 2C2H4 + Benzene + 2H2 pyr 1 ODECAL -> C3H6 + Toluene + 2H2 pyr 2 O2 + DECALIN -> HO2 + RDECAL oxi 4 O2 + DECALIN -> HO Tetralin DECALIN oxi 5 RDECAL + O2 -> RDECOO oxi 6 QDECOOH + O2 -> ZDECA oxi Sensitivity of ignition delay time Figure 5.22 Brute-force sensitivity analysis of ignition delay time of decalin in air at 20 atm (TOP) and 50 atm (BOTTOM), ϕ = 1.0. As shown in Figure 5.9, while it generally satisfactorily reproduces the ignition delay time of decalin in air at 50 atm, the original Ranzi modeling generally overpredicts the ignition delay time of decalin in air at above 900 K and 20 atm. Among the reactions shown in Figure 5.22, besides reaction oxi 1, both reactions oxi 3 and oxi 4 exhibit relatively larger sensitivities at above 900 K. One possible way of improving the performance of the original decalin oxidation sub-mechanism at higher temperatures is to multiply the rate constant of reaction oxi 4 by a factor of two, as an example. Figure 5.23 presents the results of this modification. Clearly, improved kinetic modeling is achieved for current data at 20 atm. In addition, generally good agreement is also found between previous measurements by Oehlschlaeger et al. [96] and the updated mechanism predictions using this kinetic modification, as shown in Figure Nonetheless, by no means is this modification the right approach to yielding reliable kinetics parameters. More theoretical and experimental work is needed. 91

118 T (K) Decalin/21% O 2 /N 2 Dashed lines: modified Ranzi modeling Dots: SU data t ign (ms) 1 = 0.5, 20 atm = 1.0, 50 atm = 1.0, 20 atm /T (1/K) Figure 5.23 Comparison of currently measured and newly predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 20 atm; ϕ = 1.0, 20 atm; ϕ = 1.0, 50 atm. Modified Ranzi modeling was made using the reaction mechanism [82] with modified rate constant for reaction oxi 4 in Figure 5.22 and constant U, V assumption. T (K) Decalin/21% O 2 /N 2 Dashed lines: modified Ranzi modeling Dots: RPI data t ign (ms) 0.1 = 0.5, 12 atm = 1.0, 12 atm = 1.0, 40 atm /T (1/K) Figure 5.24 Comparison of previously measured (Oehlschlaeger et al. [96]) and newly predicted ignition delay data of decalin/air mixture over K at three conditions: ϕ = 0.5, 12 atm; ϕ = 1.0, 12 atm; ϕ = 1.0, 40 atm. Modified Ranzi modeling was made using the reaction mechanism [82] with modified rate constant for reaction oxi 4 in Figure 5.22 and constant U, V assumption. 92

119 6 Chapter 6 Ignition Studies of 1-Butanol 6.1 Constrained-reaction-volume strategy The implementation and characterization of the constrained-reaction-volume approach in our low-pressure shock tube are detailed in [128]. Here, we apply this strategy in our high-pressure shock tube to achieve a near-constant-pressure autoignition environment, seeking to avoid ambiguity in data analysis that pre-ignition pressure increase entails, as mentioned in previous Chapter 3. Before we present the final pressure time-histories, both the conventional operation and staged-filling procedures will be summarized Experimental procedure In a conventional shock wave ignition-time-delay experiment, the entire driven section of the shock tube is filled with the test gas. However, in a constrained-reaction-volume experiment using staged-filling, only a small portion of the driven section is filled with test gas. Figure 6.1 shows the filling stages that the gas mixtures progress through. In the first filling stage, the test mixture (similar to the non-reactive gas except for the presence of fuel and oxidizer) is filled into the shock tube to an initial pressure (P test gas) lower than the desired ultimate driven-section pressure (P 1, total). Then, in the second stage, a non-reactive gas (in this study, N 2 or a N 2/CO 2 mixture) is filled into the driven section from a port about 200 cm away from the endwall, compressing the reactive gas from the initial-stage fill to a small slug near the endwall. To minimize mixing between the non-reactive and reactive mixtures, the second-stage filling is done slowly such that the filling takes place over two to three minutes. The length of the reactive slug at the test section before shock (L 1) can be calculated according to Eqn. 6. L 1 = (200 cm) P test gas/p 1, total Eqn. 6 The most important variable in designing a staged-filling experiment is the length of the reactive slug at the test section behind the reflected shock (L 5) [128]. This can be calculated according to Eqn. 7, where ρ 1 and ρ 5 refer to the gas densities in Regions 1 (initial static test gas) and 5 (post-reflected-shock), respectively. Empirically, we have found that the smaller the value of L 5, the more closely a constant-pressure condition can be achieved, though more dilution of the test gas can occur, due to diffusive mixing between the initial-stage test gas mixture and the second-stage N 2. In the current studies, the optimal L 5 was found to be around 4-5 cm. 93

120 L 5 = (ρ 1/ ρ 5) L 1 Eqn. 7 To ensure the accuracy and reproducibility of current CRV data, different staged-filling strategies were tested including changing the location of staged-filling port (filling at 5 m from the endwall) and using different staged-filling gas mixtures (staged-filling with N 2/CO 2 blends). Experiments performed with the N 2/CO 2 blend staged-filling gas mixtures did exhibit lower incident shock attenuation rates, but produced no statistically significant change in the measured ignition delay times. Figure 6.1 Idealized schematic of the gas mixtures during the staged-filling process Mixing characterization To establish the actual reactive gas mixtures present in the experiment, a 3.39 μm He-Ne laser was used to make quantitative measurements of the fuel concentration throughout the stagedfilling process. An example fuel concentration monitoring experiment is shown in Figure 6.2. Right after the first filling stage, the 1-butanol concentration was measured to be in good agreement (within ±10%) with the test gas mixture concentration prepared manometrically in the mixing tank. The -10% loss of fuel from 0.68% to 0.62% is attributed to wall effects in the mixing tank and shock tube. During the second filling stage, the fuel concentration oscillates somewhat owing to varying pressure. The fuel concentration decreases near the end of the second filling stage, but became relatively constant 30 s after the end of this filling stage. These experiments demonstrate that the staged-filling process has the potential to dilute the test gas mixture, so it is important to measure the fuel concentration soon after the end of the second filling stage. Additionally, it is necessary to fire the shock tube soon after the second filling stage to minimize further dilution through mixing. Figure 6.3 presents the shock-based 94

121 absorbances in Regions 1, 2 and 5 (only first 800 μs) obtained right after this monitoring experiment. The 1-butanol concentration in Region 5 (R5) was found to be approximately 10% less than that measured in Region 1 (R1) for our nominal values of L cm, owing to the shock-induced displacement of the mixture boundary. Such variation of post-shock fuel concentration is also reported in [128]. Direct fuel concentration measurements in R5 were not usually feasible in the ignition experiments because of the prohibitively high absorbances (3 4), but measurements in R2, just prior to reflected shock arrival, served to confirm this 10% reduction m Absorbance P test gas = 120 torr P 1, total = 750 torr T 1 = 91 o C X fuel, 3.39 m = 0.62% (X fuel, manometric = 0.68 %) L 1 = 32.0 cm X fuel, R1 = 0.50% 3.39 m Absorbance Pressure Fuel concentration Time (s) Pressure (torr) X fuel (%) Figure 6.2 Example monitoring of the pressure in the driven section, the 3.39 μm He-Ne absorbance and the derived 1-butanol concentration in the test gas during the staged-filling process, for staged filling with L 1 = 32 cm. X fuel,manometric, X fuel, 3.39 μm and X fuel, R1 stand for the fuel concentration prepared manometrically in the mixing tank, the one measured in situ in the shock tube at the end of first filling stage, and the one at the end of second filling stage, respectively. 95

122 3.39 m absorbance butanol/air 900 K, 16.6 atm, = 0.15 X fuel, R2 = 0.46% X fuel, R1 = 0.50% L 5 = 4.7 cm X fuel, R5 = 0.45% Time ( s) Figure 6.3 Example monitoring of the 3.39 μm He-Ne laser absorbances and the derived 1-butanol concentrations in Regions 1, 2, and 5 (X fuel, R1, X fuel, R2, and X fuel, R5) after the staged-filling process for a very lean (ϕ = 0.15) reflected shock wave experiment (shown in Figure 6.2). The slow rise immediately behind the reflected shock is associated with the reflected-shock bifurcation at the sidewall. 6.2 Conventional-filling experiments 1-butanol ignition delay times were measured at reflected-shock initial temperatures between 716 and 1121 K and nominal pressures of 20 and 40 atm at ϕ = 0.5, 1.0 and 2.0 using the conventional-filling protocol. The ignition delay time is defined as the time interval between the arrival of the reflected shock at the observation station and the onset of ignition determined by extrapolating the maximum slope of either pressure or OH* record back in time to the baseline. Ignition delay time values obtained from the pressure and emission traces were consistent within ±1%. A full listing of current test conditions and ignition delay times can be found in the Appendix A1; selected pressure traces and ignition delay time plots are given in Figure 6.4 -Figure 6.7. Ignition delay time data presented in the figures have small experiment-to-experiment pressure variations and have been scaled to a common pressure using t ign - P -1, which was experimentally validated and commonly adopted by previous researchers [ ]. The overall uncertainty in ignition delay time measurements is approximately ±20%, with the primary contribution from the uncertainty in reflected shock temperatures. 96

123 4 Relative signal (pressure) K, 22.2 atm 917 K, 21.8 atm 876 K, 15.6 atm 867 K, 16.2 atm 829 K, 17.7 atm 809 K, 17.9 atm 3.38% 1-butanol/air = 1.0 Conventional filling 792 K, 18.5 atm non-reactive Time (ms) Figure 6.4 Normalized temperature-dependent pressure profiles for ignition delay time measurements of 1-butanol in air at ϕ = 1.0 and near 20 atm with conventional-filling. Figure 6.4 shows representative pressure profiles obtained from 1-butanol ignition experiments with conventional-filling at high, intermediate and low temperatures and the influence of initial post-shock temperature on the pressure profiles and ignition delay. At the highest temperature shown (1014 K), the pressure trace is flat until a smooth and rapid exponential rise to ignition, which is usually classified as strong or sharp ignition [78, 113, 114, ]. At the lowest temperature (792 K), interestingly, the flat pressure time-history lasts for only approximately 2 ms and then slowly increases to ignition without any evidence of pressure rocketing or ringing, a form usually called mild ignition [78, 113, 114, ]. However, at the intermediate temperatures, the strong pressure rise prior to the final ignition event exhibits multiple ramps and/or humps, which can be regarded as a mild-to-strong transition ignition mode [120, 168]. These pre-ignition pressure rises are temperature dependent with lower-temperature experiments showing higher pre-ignition pressure rises; this is consistent with the findings of Vranckx et al. [103] and Lee et al. [126]. In addition, such pre-ignition effects are pressure dependent with higher-pressure experiments showing less pre-ignition pressure rise, as shown in Figure 6.5 and also reported previously [103, 126].The reasons for the pre-ignition pressure rise are discussed in later sections. 97

124 4 Relative signal (pressure) % 1-butanol/air = 1.0 Conventional filling 815 K, 34.8 atm 809 K, 17.9 atm non-reactive Time (ms) Figure 6.5 Normalized pressure-dependent pressure profiles for ignition delay time measurements of 1- butanol in air at ϕ = 1.0 with conventional-filling. The measured ignition delay times in conventional-filling experiments with ϕ = 1.0 at 20 and 40 atm are plotted in Figure 6.6 and Figure 6.7 and compared with shock tube data from previous studies [102, 107]. Non-linear Arrhenius behavior for all sets of data is observed. As shown in Figure 6.6, the current 20 atm data are in very good agreement with the previous results of Stranic et al. [107] using the same Stanford facility. However, although current data agree well with the results of Heufer et al. [102] at temperatures above 1100 K, the present data are approximately 40% lower than those of Heufer et al. [102] at temperatures below 1100 K. By contrast, and surprisingly, the present data at 40 atm are in good agreement with those measured by Heufer et al. [102] over the full temperature range studied. The reasons for such apparent inconsistency are difficult to trace, since all three sources of data collected at low temperatures were subject to pre-ignition pressure rises [102, 107] and the impact of these rises might vary between different shock tube facilities [107]. In addition, the data by Heufer et al. [102] are affected by non-reactive facility-dependent pressure gradients (dp 5/dt) of, on average, 4.5%/ms and 3.6%/ms at 20 and 40 atm, respectively, while driver inserts were employed in both the Stranic et al. work [107] and the current study to eliminate such gradients prior to ignition. Thus, it is difficult to draw clear and useful kinetic conclusions from current comparisons, other than to emphasize the role of facility-dependent effects that are not yet fully accounted for in modeling these experiments. 98

125 1250K 1000K 833K 714K 10 Heufer et al. Stranic et al. Current study t ign (ms) butanol/air, =1 20 atm Scaled by P -1 Conventional filling /T (1/K) Figure 6.6 Comparison of current ignition delay times of 1-butanol in air with previous results at ϕ = 1.0 and 20 atm with conventional-filling. 1250K 1000K 833K 714K 10 Heufer et al. Current study t ign (ms) butanol/air, =1 40 atm Scaled by P -1 Conventional filling /T (1/K) Figure 6.7 Comparison of current ignition delay times of 1-butanol in air with previous results at ϕ = 1.0 and 40 atm with conventional-filling. For the ultimate goal of validation and refinement of a 1-butanol reaction mechanism, such comparisons among different sources of data naturally raise the question as to how to accurately model conventional-filling experiments, particularly at low temperatures where substantial pre- 99

126 ignition pressure rises can exist. At the same time, there is a wish by experimentalists to advance experimental strategies that enable more quantitative ignition delay time measurements and modeling independent of uncertain thermodynamic-gasdynamic models. 6.3 Interpretation of conventional-filling ignition data Figure 6.8 shows a comparison of ϕ = 1.0 data in conventional-filling experiments and the corresponding simulation results using five different mechanisms with the CHEMKIN-PRO software suite [147] and the conventional constant-volume (V),constant-energy (U) constraints. The measured ignition delay times at intermediate temperatures ( K) are significantly shorter than all the mechanism predictions. The reasons for such discrepancies are various, but ambiguous. It could be that the mechanisms are not well tuned and optimized, or that the thermodynamic-gasdynamic constant U, V model used, is not appropriate, or both. Thus, the adjustment of reaction rate parameters to tune or refine these chemical kinetic mechanisms to fit global kinetic targets such as ignition delay time is questionable and unconvincing, a point previously stated by Lee et al. [126]. To resolve these disparities, we need to examine the pressure-time profiles during the ignition experiments and investigate the implications of gasdynamic models on simulating ignition delay times [126, 168, 169] K 1000 K 1-butanol/air 20 atm, =1 Experiment Conventional filling 833 K K 5 t ign (ms) Modeling (const. UV) 1-Grana et al. (2010) 2-Black et. al. (2010) 3-Merchant et al. (2012) 4-Sarathy et al. (2012) 5-Vranckx et al. (2011) /T (1/K) Figure 6.8 Comparison of current conventional-filling ignition delay times of 1-butanol in air at ϕ = 1.0 and 20 atm with modeling results using five recent mechanisms [103, ]. 100

127 Normalized PCB signal % 1-butanol/air 876 K, =1, 15.6 atm Conventional filling Kistler Time (ms) Figure 6.9 Normalized pressure traces at five measurement locations for a 1-butanol/air mixture at ϕ = 1.0 with conventional shock tube operation. In previous studies [78, 113, 119, , 166], pressure traces recorded from transducers mounted at several locations on the driven section were utilized to unravel evidence of a remote (i.e., outside the reaction volume near the endwall) ignition phenomenon that occurred with long ignition delay times. A similar analysis was conducted here. Figure 6.9 presents the measured pressure time-histories at each of the five axial locations along the driven section as well as a duplicate pressure measurement at the 1.1 cm location recorded using a Kistler pressure transducer. We can see that the Kistler, PCB 5 and PCB 4 all exhibit features of pre-ignition pressure rises. Specifically, such rise appears to first take place at the Kistler location and then propagates to PCB 5 and PCB 4. Since the Kistler transducer is located at the observation point for the ignition delay time measurements, this observation favors the point of view that ignition in this case starts near endwall. This shock-induced inhomogeneous ignition process is characterized by a deflagration-to-detonation transition (DDT) of the fuel/air mixture where a moderate ramp is followed by a steep rise. Significant increases in pressure due to the development of this detonation are observed subsequently in the recordings of PCB 3, PCB 2 and PCB 1. In this specific experiment, then, the pre-ignition pressure change is likely not due to a distant or remote ignition, but could be due to local inhomogeneities in the reaction volume behind the reflected shock, or merely a result of pre-ignition chemistry. 101

128 The causes of such inhomogeneities, however, are difficult to ascertain. Fiewegeret al. [78] point out that these might manifest themselves as non-uniformities in temperature, radical concentrations, or particles in the reflected shock region. Dryer et al. [170] discuss the potential relevance of inhomogeneous mixing and catalytic processes as perturbing factors. In an earlier study in our same shock tube facility, Petersen et al. [130] specifically discuss the causes of inhomogeneities by facility-dependent effects such as shock-boundary layer interaction. In another shock tube facility, Shen et al. [171] found that the presence of contaminants in the reactants or on experimental surfaces (soot deposits) in their facilities contributed to the occurrence of pre-ignition ramp behavior; however, this explanation was disproved by Vasu et al. [120] for the Stanford facility used in current work. More recently, Heufer et al. [127] adds that inhomogeneous ignition may also be strongly influenced by fuel chemistry. We reserve further investigation of the real causes of pre-ignition pressure-rise behavior for future work. A more important current task is to accurately model these pre-ignition-affected experiments. 80 Pressure (atm) % 1-butanol/air 876 K, =1, 15.6 atm Measured P profile ramp 2 ramp 1 Conventional filling Specified P Simulated P (Sarathy et al. constant. UV) non-reactive Time (ms) Figure 6.10 Measured, specified, and simulated pressure profiles for an example ignition delay time measurement with conventional-filling. The simulated pressure profile was from calculations using the reaction mechanism of Sarathy et al. [111] and a constant U, V assumption. In a preliminary effort to study this problem, two thermodynamic-gasdynamic models are discussed here. The first one is the traditional constant U, V model. The validity of such a model can be checked by comparing measured and simulated pressure profiles [143], as shown in Figure The simulations in this study were performed using the reaction mechanism of Sarathy et al. [111]. Interestingly, a pre-ignition pressure ramp is evident in both the 102

129 measurement and simulation. However, a closer examination reveals that the measured ramp is composed of a slow exponential rise (ramp 1) followed by a fast increase (ramp 2) to the time of ignition. The simulation, however, can only reproduce ramp 1, but not ramp 2, which supports the thesis that constant U, V is not suitable for the simulation of such experimental data. The second thermodynamic-gasdynamic model is to use a specified pressure profile and solve the energy equation. This method has been used previously by others including Aul et al. [169]. In this model, the measured reflected-shock pressure profile up to the instant of ignition is input directly into a zero-dimensional homogeneous batch reactor model to constrain the pressure; this then fixes the time-history of the enthalpy for an adiabatic system. We will refer to this as the specified-pressure method. An example specified-pressure profile is shown in Figure In these simulations, we directly used our pressure data together with the specified P option in CHEMKIN-PRO to solve for the OH mole-fraction time-history and hence ignition delay time. The same procedure was repeated for the other data affected by pre-ignition pressure change. We are aware that some researchers do not follow this procedure. For example, Chaos et al. [172] and Lee et al. [126] make use of pressure time-history data to generate time-varying volume and temperature profiles, respectively, based on an isentropic assumption, and used these profiles in predicting ignition delay times for comparison with data. However, this approach still communicates the pre-ignition pressure-rise information to the modeling K K 1-butanol/air 20 atm, =1 Conventional filling Current study 833 K 714 K t ign (ms) Sarathy et al. Const. UV Sarathy et al. Specified P Vranckx et al. Const. UV Vranckx et al. Specified P /T (1/K) Figure 6.11 Comparison of current conventional-filling ignition delay data with two model predictions using both constant U, V assumption and the specified-pressure method. 103

130 Figure 6.11 presents a comparison of the current conventional-filling data with two mechanism predictions using two kinds of gasdynamic models, i.e., constant U, V and specified-pressure models, respectively. At higher temperatures, as shown in Figure 6.4 at 1014 K, there is no evidence of pre-ignition ramps, so the traditional constant U, V model is a good approximation of the actual test gas state. However, at lower temperatures where pre-ignition pressure change is observed, significant improvement of modeling results can be achieved by using the specifiedpressure method. This is true for the three mechanism predictions investigated here [103, 111, 112]. Nonetheless, a caveat must be offered that such agreement is artificial in that the specifiedpressure method tends to force ignition near the measured induction time. This may explain why the three mechanisms yielding significantly different values for constant U, V collapse to near-identical predictions when incorporating the specified-pressure time-histories. Thus, a satisfactory interpretation of shock tube experiments containing pre-ignition pressure increases is still lacking, and more theoretical and computational work is needed to predict the trends of pressure change shown in Figure Constrained-reaction-volume experiments As mentioned above, in order to obviate misinterpretation of ignition delay data, we applied a newly developed and validated driven-gas loading strategy, a constrained-reaction-volume (CRV) strategy [128], to the 1-butanol ignition measurements. Results of the experiments using this strategy at both high and low temperatures are presented below. Figure 6.12 shows two sample measurements for the higher temperature case of 1047 K with both the conventional-filling and CRV approach. The ignition delay times in both experiments are very similar, as expected, particularly for our observation station so close to the endwall (1.1 cm). Additionally, no clear evidence of pre-ignition energy release is observed at this temperature. This is understandable, as in the strong ignition regime discussed given earlier, a fairly sharp, uniform combustion wave starts at the endwall and then propagates into the test gas. This strong ignition mode has been seen in many optical visualization studies in shock tubes [78, 127, 166, 173]. It also should be noted that prior to ignition, as indicated by the steepest rise of the pressure, the pressure is fairly constant, which implies that using a constant-pressure constraint is a valid gasdynamic model for the simulation of these experiments up to ignition. 104

131 Pressure (atm) Pressure (non-reactive) Conventional filling Pressure (reactive) 306 nm OH* Emission CRV (L 5 = 4 cm) Pressure (reactive) 306 nm OH* Emission 3.38% 1-butanol/air 1047 K, 16.3 atm = Time ( s) 4 3 OH* emission signal 2 1 Figure 6.12 Comparison of conventional-filling and CRV ignition delay data at 1047 K, -20 atm, ϕ = 1.0. However, at lower temperatures, for data affected by pre-ignition energy release, measurements using the CRV strategy exhibit a dramatic difference, as shown in Figure The pressure in the CRV experiment is reasonably constant, except a tiny bump from 3-7 ms that is likely due to imperfect tailoring of the driver gas. In the CRV experiment, the energy release during chemical induction dissipates to the neighboring non-reactive gas (N 2 in this case) instead of creating a pre-ignition pressure ramp. This dissipation also explains the absence of any detonation-like pressure ringing. In the CRV experiments, the onset of ignition is marked by a rise in OH* emission. Not only is the ignition delay time in the CRV experiment now longer than that in the conventional-filling one, but the peak of the emission trace is lower. Such phenomena are expected in the CRV experiment as the pressure change (and concomitant temperature change) during the ignition process is far less than that observed in the conventional-filling one. More importantly, an examination of the pressure traces at various axial locations (see Figure 6.9) confirms that remote ignition did not occur. Thus, with the absence of either pre-ignition pressure change or remote ignition, the CRV approach enables the unambiguous and reliable quantitative modeling of ignition delay time using specified P, H gasdynamic constraints. Pressure histories even closer to constant P could be achieved with lower fuel concentrations and/or smaller L

132 Pressure (atm) Pressure (non-reactive) Pressure (reactive) 306 nm OH* Emission Conventional filling 3.38% 1-butanol/air 876 K, =1, 15.6 atm OH* emission signal Pressure (atm) Pressure (non-reactive) Pressure (reactive) 306 nm OH* Emission 3.38% 1-butanol/air 882 K, =1, 17.5 atm L 5 = 4 cm CRV strategy 6 OH* emission signal Time (ms) Time (ms) Figure 6.13 Comparison of conventional-filling(left) and CRV(RIGHT) ignition delay data at -880 K, -20 atm, ϕ = 1.0. More examples of CRV experiments at different temperatures and oxygen concentrations are shown in Figure For the same reasons mentioned above, the OH* emission peaks decrease with decreasing temperature. Note the very high temperature sensitivity of the OH* histories. In future work, we anticipate using laser absorption to measure OH and temperature, providing important new target data for quantitative comparisons with detailed kinetics simulations based on the specified P, H gasdynamic model. 306 nm OH* emission signal Pressure (reactive) Pressure (non-reactive) 3.38% 1-butanol/air =1, ~ 20 atm L 5 = 4 cm 882K 869K 850K CRV strategy 848K 839K Time (ms) Pressure (atm) 306 nm OH* emission signal Pressure (non-reactive) Pressure (reactive) 874K 851K 822K 3.38% 1-butanol/40.6% O 2 /N 2 =0.5, ~ 20 atm L 5 = 4 cm CRV strategy 800K 777K Time (ms) Pressure (atm) Figure 6.14 Example CRV measurements of ignition delay time at low temperatures for 3.38% 1- butanol/20.3% O 2/N 2 (LEFT) and 3.38% 1-butanol/40.6% O 2/N 2 (RIGHT). Figure 6.15 shows a comparison of data from both conventional-filling and CRV experiments at ϕ = 1.0 and 20 atm. At temperatures above ~1000 K, both types of experiments yield consistent data, while at temperatures below ~1000 K, the data in conventional-filling experiments become increasingly shorter than those in CRV experiments. Similar phenomena 106

133 are also observed for data at ϕ = 0.5 and 2.0 (not shown here, but included in the Appendix A1). Moreover, while the conventional-filling data show weak evidence of NTC, the CRV data show nearly no evidence of NTC. However, as discussed in Section 6.3, conclusions about 1-butanol low-temperature-chemistry cannot reliably be drawn from the conventional-filling data. This comparative study is, to the best of our knowledge, the first of its type. 1250K K CRV Strategy Conventional Filling 833K 714K t ign (ms) 1 1-butanol/air 20 atm, = 1.0 All data scaled by P /T (1/K) 1.4 Figure 6.15 Comparison of conventional-filling and CRV ignition delay times of 1-butanol in air at ϕ = 1.0 and 20 atm. Solid lines are simply best fits to data. 6.5 Interpretation of CRV ignition data The use of the CRV strategy has allowed acquisition of an improved ignition delay time dataset with a well-defined, near-constant-pressure gasdynamic state. A clear and useful comparison of data and modeling is now possible. Figure 6.16 shows a comparison of CRV data at ϕ=1.0 and corresponding modeling results using three recent mechanisms and a constant P, H gasdynamic assumption. The Merchant et al. mechanism [112] overpredicts the ignition delay time of 1- butanol at low temperatures by a factor of 4-5. The Sarathy et al. [111] mechanism shows excellent agreement with all CRV data. The Vranckx et al. [103] mechanism also shows excellent agreement with all CRV data, except at the lowest temperatures where the mechanism predictions show evidence of an NTC behavior at K. This might be due to the addition of a simplified butyl peroxy chemistry, which was mainly based on ethanol kinetics, in an effort to fit their 1-butanol ignition delay data at ϕ = 1.0 and 80 atm [103]. However, pre-ignition pressure rises are still evident in that study [103], so it remains uncertain whether those rises are 107

134 due to genuine fuel chemistry or not. In contrast, Sarathy et al. [111] state that low-temperature reactivity is suppressed in their model due to a choice of a reaction rate parameter for the reaction 1-hydroxybutyl + O 2 = nc 3H 7CHO + HO 2 that allows it to compete with the lowtemperature chain-branching reactions. We attempted to conduct CRV experiments at temperatures below 820 K, but we did not observe any ignition signal within our 10-ms test time limit. Hence, we have not yet obtained any CRV data below 820 K. However, extension of our test times is planned that should provide valuable information to confirm or contradict the existence of NTC behavior at temperatures below 820 K. t ign (ms) 1250K K Modeling (const. HP) 1-Merchant et al. 2-Sarathy et al. 3-Vranckx et al. Experiment CRV data 833K 1-butanol/air 20 atm, =1.0 All data scaled by P K /T (1/K) Figure 6.16 Comparison of current CRV data at ϕ= 1.0, 20 atm with model predictions at constant H, P. To avoid the complications introduced by pre-ignition pressure perturbations, the CRV strategy was applied in the investigation of equivalence ratio and oxygen concentration dependence of 1-butanol ignition delay time at low temperatures. Figure 6.17 and Figure 6.18 present comparisons of current CRV data at 20 atm and Sarathy et al. mechanism predictions at different equivalence ratios/oxygen concentrations. As shown in Figure 6.17, the overall agreement between the trends of the data and modeling is good; there is general agreement in terms of the magnitude of ignition delay data and the shape of the general temperature-dependent trend. The experimental variation with equivalence ratio at fixed oxygen concentration, however, is smaller than in the simulation. Agreement is good for the rich mixtures in the case of fixed fuel concentration (3.38%) and the 10.2 and 20.3% oxygen concentrations, as shown in Figure However, the simulations fail to predict the strong oxygen concentration dependence between 108

135 20.3% % O 2. Modeling results using two other mechanisms [111, 112] (not shown) also fail to predict this strong oxygen concentration dependence. t ign (ms) K K CRV data modeling const. HP 833K =0.5 1-butanol/air =1.0 = K 20 atm All data scaled by P /T (1/K) Figure 6.17 Comparison of current CRV data and Sarathy et al. [111] mechanism predictions at ϕ = 0.5, 1.0, 2.0, 20 atm with oxygen concentration being fixed (-20%) K K CRV data modeling const. HP 833K 714K t ign (ms) % O 2 = % O 2 = % 1-butanol/O 2 /N % O 2 = atm All data scaled by P /T (1/K) Figure 6.18 Comparison of current CRV data and Sarathy et al. [111] mechanism predictions at ϕ = 0.5, 1.0, 2.0, 20 atm with fuel concentration being fixed (3.38%). 109

136 110

137 7 Chapter 7 Summary and Future Work 7.1 Engine fuels Ignition delay times for a wide variety of distillate fuels could be well correlated if care was taken to identify the important variables and areas of commonality. Of importance is the observation that ignition delay times for distillate petroleum fuels in air with stoichiometries near unity scale together when pressure and temperature are considered for temperatures above 1000 K. Large variations in fuel composition do not appear to significantly affect the ignition delay times. This observation is useful in supporting the approach of using a common detailed or reduced mechanism to describe high-temperature jet fuel oxidation. Equally interesting is the observation that at temperatures above 1000 K, there is only a weak dependence of ignition delay time on equivalence ratio in the range from 0.85 to There does appear to be a crossover temperature near 1200 K, where below this temperature the ignition shortens with higher equivalence ratio and above this temperature the ignition time lengthens. Additionally, the activation energy of the ignition delay time was also found to be dependent on oxygen concentration for equivalence ratios near unity, with lower concentrations experiencing higher activation energies. Besides ignition delay time, a database of pyrolysis product species time-histories was generated for three FAA Category-A fuels, six FAA Category-C fuels, and two RP-2 fuels. These laser absorption data are characterized by very accurate and uniform test conditions, fuel loading and species concentrations. Measured ethylene and methane product yields were found to be similar for all three Category-A fuels over temperatures from 1100 to 1350 K at 12 atm. While ethylene yields appear to plateau or peak at higher end temperatures, methane yields seem to keep increasing at current temperature range. These ethylene and methane data provide kinetic targets for engine fuel kinetics models that use product yields to constrain reduced pyrolysis sub-mechanisms (of the HyChem model). The HyChem model needs more refinement, and more data will be collected for this data-and-model iteration and optimization. 7.2 Decalin Ignition delay time measurements and 3.39 and 10.6 μm diagnostics were performed during the oxidation and pyrolysis of decalin, respectively, over a wide range of temperatures, pressures and equivalence ratios. The measured ignition delay data are characterized by relatively low 111

138 scatter and strong variation with pressures and equivalence ratios. For temperatures below 920 K, current ignition delay measurements provide clear evidence of pre-ignition heat release and NTC roll-off behavior. While comparisons of previous and current measurements yield the same pressure dependence but different equivalence ratio dependence, the chemical kinetic predictions of decalin/air ignition delay time are generally in fairly good agreement with current data. In addition, the 10.6 μm diagnostic provides unique and critical target data, i.e., ethylene concentration time-histories, for improvement of the decalin pyrolysis system. Adjustment of one of the oxidation reaction rate coefficient by a factor of two yields better agreement between experiment and modeling. Besides current preliminary modifications for oxidation kinetics of decalin, we believe that the adjustment of decalin reaction mechanism requires further theoretical and experimental work Butanol Ignition delay time measurements were carried out at elevated pressures and low temperatures for 1-butanol/O 2/N 2 mixtures at various equivalence ratios/oxygen concentrations using both conventional-filling and staged-filling strategies. Conventional-filling experiments with preignition pressure perturbations were inherently difficult to interpret and it is uncertain which gasdynamic model is the most appropriate to use in simulations. To resolve this problem, we applied the CRV concept and showed that this new strategy enables the generation of a wide range of ignition delay data that can be modeled effectively using a constant or specified P, H gasdynamic assumption. Using this CRV strategy, more confident comparisons of data with simulations from existing 1-butanol detailed reaction mechanisms are now possible. The current 1-butanol ignition delay time data does not show NTC behavior at the conditions of this study, in agreement with several of the models tested. We expect that besides the advantage of offering accurate constant P, H target data at low temperatures for modeling, the CRV concept will allow a view of the entire oxidation process via species and temperature time-history measurements from fuel decomposition, through ignition, to final products, thereby providing more extensively meaningful datasets in evaluating and refining detailed kinetic mechanisms. 112

139 8 Appendices 8.1 A1: Tables of Raw IDT Data Tables of raw (unscaled) IDT for fuels tested in current studies are listed in Table 8.1 -Table 8.34 below. Table 8.1 Ignition Delay Times of A1 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air

140

141 Table 8.2 Ignition Delay Times of A2 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air Table 8.3 Ignition Delay Times of A3 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air

142 Table 8.4 Ignition Delay Times of C1 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air

143 Table 8.5 Ignition Delay Times of C5 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air Table 8.6 Ignition Delay Times of R4 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air

144 Table 8.7 Ignition Delay Times of R5 Bath tign (µs) T (K) P (atm) ϕ 4%O2/Ar Air Table 8.8 Ignition Delay Times of C2 Bath tign (µs) T (K) P (atm) ϕ Air Table 8.9 Ignition Delay Times of C3 Bath tign (µs) T (K) P (atm) ϕ Air

145 Table 8.10 Ignition Delay Times of C4 Bath tign (µs) T (K) P (atm) ϕ Air Table 8.11 Ignition Delay Times of C6 Bath tign (µs) T (K) P (atm) ϕ Air Table 8.12 Ignition Delay Times of J11 P (atm) ϕ T (K) tign (μs) -3atm, ϕ-1.0 in air atm, ϕ atm, ϕ atm, ϕ

146 Table 8.13 Ignition Delay Times of C7 P (atm) ϕ T (K) tign (μs) -3atm, ϕ-1.0 in air atm, ϕ atm, ϕ atm, ϕ

147 Table 8.14 Ignition Delay Times of J11/C7 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -3atm, ϕ-1.0 in air atm, ϕ atm, ϕ atm, ϕ

148 Table 8.15 Ignition Delay Times of B12 P (atm) ϕ T (K) tign (μs) -3atm, ϕ-1.0 in air atm, ϕ atm, ϕ atm, ϕ

149 Table 8.16 Ignition Delay Times of J11/B12 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -3atm, ϕ-1.0 in air atm, ϕ atm, ϕ atm, ϕ Table 8.17 Ignition Delay Times of S8 P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air Table 8.18 Ignition Delay Times of S9 P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air

150 Table 8.19 Ignition Delay Times of J11/S8 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air Table 8.20 Ignition Delay Times of J11/S9 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air Table 8.21 Ignition Delay Times of H10 P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air

151 Table 8.22 Ignition Delay Times of J11/H10 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air Table 8.23 Ignition Delay Times of H11 P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air Table 8.24 Ignition Delay Times of J11/H11 blend (50:50 v.%) P (atm) ϕ T (K) tign (μs) -6 atm, ϕ-1.0 in air

152 Table 8.25 Ignition Delay Times of F15 P (atm) ϕ T (K) tign (μs) -17 atm, ϕ-0.26 in 4%O2/Ar atm, ϕ-0.54 in 4%O2/Ar atm, ϕ-1.0 in 4%O2/Ar atm, ϕ-1.2 in 4%O2/Ar atm, ϕ-1.0 in 4%O2/Ar

153 Table 8.26 Ignition Delay Times of H13 P (atm) ϕ T (K) tign (μs) -17 atm, ϕ-1.0 in 4%O2/Ar Table 8.27 Ignition Delay Times of B14 P (atm) ϕ T (K) tign (μs) -17 atm, ϕ-1.0 in 4%O2/Ar Table 8.28 Ignition Delay Times of B15 P (atm) ϕ T (K) tign (μs) -17 atm, ϕ-1.0 in 4%O2/Ar

154 Table 8.29 Ignition Delay Times of HRA P (atm) ϕ T (K) tign (μs) -17 atm, ϕ<1.0 in 4%O2/Ar atm, ϕ-1.0 in 4%O2/Ar atm, ϕ>1.0 in 4%O2/Ar atm, ϕ-1.0 in 4%O2/Ar

155 Table 8.30 Ignition Delay Times of K6 tign (µs) T (K) P (atm) P average ϕ Xfuel XO2 XN2 ϕ bin

156

157 Table 8.31 Ignition Delay Times of K7 tign (µs) T (K) P (atm) P average ϕ Xfuel XO2 XN2 ϕ bin

158 Table 8.32 Ignition Delay Times of K8 tign (µs) T (K) P (atm) P average ϕ Xfuel XO2 XN2 ϕ bin

159 Table 8.33 Ignition Delay Times of decalin in air P (atm) ϕ T (K) tign (μs) -20 atm, ϕ atm, ϕ atm, ϕ

160 atm, ϕ atm, ϕ atm, ϕ

161 Table 8.34 Ignition Delay Times of 1-butanol/O 2/N 2 T5 (K) P5 (atm) tign (μs) 3.38% 1-butanol/20.3% O2/N2, ϕ=1.0; conventional-filling % 1-butanol/20.3% O2/N2, ϕ=1.0; CRV, L5-4 cm % 1-butanol/20.6% O2/N2, ϕ=0.5; conventional-filling % 1-butanol/20.6% O2/N2, ϕ=0.5; CRV, L5-4 cm % 1-butanol/19.6% O2/N2, ϕ=2.0; conventional-filling % 1-butanol/19.6% O2/N2, ϕ=2.0; CRV, L5-4 cm 135

162 % 1-butanol/40.6% O2/N2, ϕ=0.5; conventional-filling % 1-butanol/40.6% O2/N2, ϕ=0.5; CRV, L5-4 cm % 1-butanol/10.2% O2/N2, ϕ=2.0; conventional-filling % 1-butanol/10.2% O2/N2, ϕ=2.0; CRV, L5-4 cm % 1-butanol/20.3% O2/N2, ϕ=1.0; conventional-filling

163 8.2 A2: Tables of Product Yields Table 8.35 Product yields during pyrolysis of A1, A2, and A3 (SK3 dataset) Run # Fuel (%) T5 P5 C2H4 yield CH4 yield C2H4 yield CH4 yield (K) (atm) 0.5 ms 0.5 ms 1.0 ms 1.0 ms C2H4 yield 1.5 ms CH4 yield 1.5 ms A A A

164 Table 8.36 Product yields during pyrolysis of C1, C2, C3, C4, C5, and C6 Fuel Fuel (%) T5 P5 C2H4 yield (K) (atm) 0.5 ms CH4 yield 0.5 ms C C C C C C

165 Table 8.37 Product yields during pyrolysis of R4 and R5 Fuel T5 P5 Fuel C2H4 CH4 C2H4 yield CH4 yield C2H4 yield CH4 yield C2H4 yield (K) (atm) (%) data data 0.5 ms 0.5 ms 1.0 ms 1.0 ms 1.5 ms R yes yes CH4 yield 1.5 ms yes yes yes yes yes yes yes no R yes yes yes yes yes yes yes no yes no yes no yes no

166 8.3 A3: 3.39 µm Absorption Cross Sections of Fuels The absorption cross section of several test fuels were collected at 3.39 µm and temperatures of K, as shown in Figure 8.1 -Figure 8.4. These spectroscopic data provide valuable information for quantitative detection of hydrocarbon fuels at a wide range of temperatures. Cross Section (m 2 /mol) m B12 C7 J11/C7 (50:50 v.%) J T (K) Figure 8.1 Temperature-dependent absorption cross section of B12, C7, J11, and J11/C7 Blend at 3.39 μm. 100 Cross Section (m 2 /mol) m HPST & He-Ne B14 F15 H13 B T (K) Figure 8.2 Temperature-dependent absorption cross section of B14, F15, H13, and B15 at 3.39 μm. 140

167 60 Cross Section at 3.39 m (m 2 /mol) A1 A2 A T (K) Figure 8.3 Temperature-dependent absorption cross section of A1, A2, and A3 at 3.39 μm. Cross Section (m 2 /mol) Decalin (g) 3.39 m (2948 cm -1 ) HPST & He-Ne FTIR Best fit T (K) Figure 8.4 Temperature-dependent absorption cross section of decalin at 3.39 μm. Representative error bars are shown. 141

168 8.4 A4: Shock Attenuation Issue in HPST CRV Introduction In previous constrained-reaction-volume (CRV) experiments for ignition delay time of 1- butanol in air on the high-pressure shock-tube (HPST), pure N 2 was used as the staged-filling gas and was added from a port around 200 cm from the driven section endwall. Surprisingly, the shock attenuation rate was very high (7 8 %/m) for low-temperature measurements (below 900 K). Although the last three or even two velocity points were employed to achieve quite low uncertainty of velocity fitting (less than 0.3%), this high attenuation rate is still worrisome in that the calculations of shock conditions become very sensitive to how we extrapolate the velocity profile to the endwall. Since T 5 and P 5 are crucial in shock tube experiments, we want to conduct new CRV experiments with low attenuation rate to test the reliability of old CRV data. In this report, a simple theoretical analysis about interaction between shock wave and contact surface is illustrated first providing some guidance for new CRV experiments. Then, results of attenuation rates from new CRV experiments are presented and discussed. Finally, both old and new CRV data are compared with mechanism predictions Theory A generic analysis is shown here concerning the interaction between a shock wave and a contact surface separating two different gases (gas 1 and gas 1 ). A shock wave propagates from Region 1 to Region 1, both of which are at rest initially with the same initial temperature and pressure. We know from the shock-jump relations that, across the shock, u a ( P 1) ( P 1) a 2 P21 ( 1 P21 ) a1 1P (1) (2) 142

169 where 2, 1 1( 1 1) If an expansion wave is reflected from the contact surface, we have, across the expansion wave, where u u a ( 1 P 1) 32 (3) Notice that u3 u2 u3 u 2 a 2 a1 a1 a2 a1 Substitute eqns. (1) (3) into eqn. (4), we get (4) u3 1( P21 1) ( 1 21) 0.5 (1 P P P32 ) (5) a1 ( 1P21 1) 1 1 1P21 1 There should be a transmitted shock wave passing through the contact surface, so across Region 1 and 2 we have 0.5 u a ( P 1) ( P 1) 2' 1' 2'1' 1' 1' 2'1' Across the contact surface we must have u3 u2', P3 P2'. Combining eqns. (5) and (6), we derive that 0.5 (6) a1' 1 P21 1 P 1 21( 1 P21 ) ( 1' P31' 1) (1 P ) (7) a1 1' ( 1P21 1) 1( 1P21 1) P31' 1 Now we assume that no reflection of waves would occur, i.e., P32 1. So eqn. (7) becomes or a a ( P 1) ( P 1) 1' 1 1' ' (8) 143

170 1' 1 1' 1 ( P21 1) a 2 1' 1' a ( P21 1) 2 1 Comparing eqns. (7) and (9), we conclude that if (9) 1' 1 1' 1 ( P21 1) a 2 1' 1' (10) 0.5 a ( P21 1) 2 1 an expansion wave will be reflected from the contact surface and the transmitted shock will be faster than the incident one, and vice versa Experiments Current new CRV experiments consist of two successive parts. First, we changed the stagedfilling location to see whether it makes any difference in the attenuation rate and ignition delay time. Then, we tried to match the staged-filling gas and test gas based on the theory discussed above for reduction of attenuation rate. Results with respect to attenuation rate will be reported and discussed first, followed by effects of these changes in staged-filling strategy on the ignition delay time of 3.38% 1-butanol/air (test gas) at -20 atm Change the staged-filling location Figure 8.5 shows comparison of velocity measurements from two different staged-filling locations, i.e., 200 cm and 496 cm from endwall, respectively. The staged-filling gas was still pure N 2. Evidently, not only do they show similar pattern, such as change of slope at V 2, but also the overall attenuation rates in both experiments appear the same. In addition, the fuel loss caused by the diffusive mixing in both cases is around 40%. Therefore, we conclude that moving the staged-filling location from middle port to near the diaphragm does not make a difference in the attenuation rate or the extent of diffusive mixing for the same L 1. The large attenuation rate observed in these experiments, especially the one associated with V 2 V 4, can be explained by comparing the acoustic impedance ( a ) of pure N 2 and test gas, as shown in Table The acoustic impedance of test gas is larger than that of pure N 2 so that the incident shock slows down even further through the broad mixing zone between the staged-filling gas and test gas

171 The velocity points used for fitting and the velocity fits are also presented in Figure 8.5. The large attenuation rate associated with these fits makes the calculations of T 5 and P 5 highly sensitive to the velocity extrapolated to the endwall. Velocity (m/s) Staged-filling from near the diaphragm 496 cm from endwall Staged-filling from the middle port 200 cm from endwall shock propagation direction V 4 V 3 V 2 V 1 test gas: 1-butanol/air staged-filling gas: pure N 2 L 1 = 30 cm Distance from endwall (cm) Figure 8.5 Comparison of velocity measurements from two different staged-filling locations, i.e., 200 cm and 496 cm from driven section endwall, respectively Table 8.38 Properties of staged-filling gas (gas 1) and test gas (gas 1 ) at 1 atm and 91 o C MW: molecular weight γ: specific heat ratio 145

172 a : speed of sound ρ: density a : acoustic impedance Change the staged-filling gas Figure 8.6 plots the left-hand-side (LHS) and right-hand-side (RHS) of Eqn. (9), where gas 1 refers to a mixture of CO 2 and N 2 and gas 1 test gas, as a function of CO 2 concentration. In theory, to satisfy Eqn. (9), 11.6% CO 2 should be chosen. Another way of understanding this choice is, again, comparing the acoustic impedance ( a ) of staged-filling gas and test gas, as shown in Table The acoustic impedance of 11.6% CO 2/N 2 is nearly the same as that of test gas, which implies that with this choice the contact surface between staged-filling gas and test gas does not affect the propagation of incident shock so that the attenuation rate should be as normal as that in conventional-filling experiments (1 2 %/m). However, incidentally, the aforementioned broad mixing zone occupies several stations of velocity measurement (see the schematic of V 1 V 4 in Figure 8.5), and the concentration of CO 2 in the mixing zone might be lower than expected. Therefore, a mixture of CO 2 and N 2 with initially 11.6% CO 2 might not be effective in ameliorating the attenuation rate in the last 123-cm-long section of the driven section. Mixtures with initial concentrations of 20% and 30% CO 2 are also tested. 1.3 LHS or RHS of Eqn Gas 1: CO 2 /N 2 Gas 1': 3.38% 1-butanol/air P 1 = P 1' = 1 atm T 1 = T 1' = 91 o C LHS RHS (P 21 = 5) X CO2 in CO 2 /N 2 Figure 8.6 Left-hand-side (LHS) and right-hand-side (RHS) of Eqn. (9) 146

173 Figure 8.7 shows comparison of velocity measurements for various concentrations of CO 2 in the staged-filling gas mixture of CO 2 and N 2. Clearly, adding CO 2 significantly helps reduce the overall attenuation rate with the higher initial concentration of CO 2 yielding the lower attenuation rate. Such variation of attenuation rate is summarized and plotted in Figure 8.8. In addition, the general pattern of velocity measurements with CO 2 added appears more linear than that with pure N 2. Figure 8.7 also presents the velocity fits extrapolated to the endwall for four cases. The uncertainty of T 5 in CRV experiments with lower attenuation rate should be lower than that with higher attenuation rate. 880 All staged-filling from near the diaphragm shock propagation direction 11.6% CO 2 Velocity (m/s) % CO 2 0% CO 2 30% CO 2 test gas: 1-butanol/air staged-filling gas: N 2 /CO 2 L 1 = 30 cm Distance from endwall (cm) Figure 8.7 Comparison of velocity measurements for various concentrations of CO 2 in the staged-filling gas mixture of CO 2 and N 2 147

174 9 Staged-filling from near the diaphragm Staged-filing from the middle port Attenuation rate (%/m) X CO2 in CO 2 /N 2 Figure 8.8 Variation of overall attenuation rate as a function of initial concentration of CO 2 in the stagedfilling gas Effects of changing staged-filling strategy on ignition delay Figure 8.9 shows comparison of ignition delay times from the experiments mentioned above and previous ones (pure N 2 staged-filled from the middle port). While the ignition delay time from the experiment with pure N 2 staged-filled from near the diaphragm is in good agreement with previous data, those from experiments with CO 2 added in the staged-filling gas appear to be slightly longer than those in previous experiments at the same conditions. Nonetheless, given that the uncertainties of T 5 and ignition delay time associated with previous experiments are relatively bigger than those in aforementioned new CRV experiments, as reflected by the horizontal and vertical error bars, the agreement is still satisfactory. 148

175 25 From middle w/ pure N 2 20 From diaphragm w/ pure N 2 909K 833K 769K 15 From diaphragm w/ 11.6% CO 2 /N 2 From diaphragm w/ 20% CO 2 /N 2 t ign (ms) 10 From diaphragm w/ 30% CO 2 /N 2 CRV 5 1-butanol/air 20 atm, = 1.0 All data scaled by P /T (1/K) Figure 8.9 Comparison of ignition delay times from new and old CRV experiments Finally, both the old and new CRV data are plotted together and compared with kinetic modeling using three recent reaction mechanisms for 1-butanol, as shown in Figure As concluded before with only old CRV data, kinetic predictions from mechanisms by Sarathy et al. and Vranckx et al. are in good agreement with all CRV data at current experimental conditions, while those by Merchant et al. generally overpredict the ignition delay times of 1-butanol/air at -20atm. t ign (ms) 1250K K Modeling (const. HP) 1-Merchant et al. 2-Sarathy et al. 3-Vranckx et al. Experiment Old CRV data New CRV data 833K 1-butanol/air 20 atm, =1.0 All data scaled by P K /T (1/K) Figure 8.10 Comparison of all CRV data and kinetic predictions using three reaction mechanisms Conclusions and Future Work The shock attenuation problem observed in previous CRV experiments is treated both theoretically and experimentally. We conclude that 149

176 1) the staged-filling location has no effect on the attenuation rate; 2) matching the acoustic impedance of staged-filling gas and test gas (by addition of CO 2 into staged-filling gas in this study) can help reduce the attenuation rate; 3) new CRV data are in fairly good agreement with old ones. Through this study, we have the following suggestions for future CRV experiments on the HPST. 1) separate studies of T 5 measurement with both high and low attenuation rate to test the validity of conventional method of obtaining velocity at the endwall (fitting and extrapolation); 2) simultaneous measurement of temperature time-history if possible, such as using singleended laser absorption; 3) staged-filling from near the diaphragm through multiple holes that are radially evenly distributed in the shock tube, with simultaneous monitoring of initial fuel concentration. 150

177 8.5 A5: Fuel Specifications Fuel specifications listed here of Category-A jet fuels, Category-C jet fuels, and AF RP-2 fuels are directly received from the end-users [60] (see permission letter below). 151