THE DEVELOPMENT AND APPLICATION OF AEROSOL SHOCK TUBE METHODS FOR THE STUDY OF LOW-VAPOR-PRESSURE FUELS

Transcription

1 THE DEVELOPMENT AND APPLICATION OF AEROSOL SHOCK TUBE METHODS FOR THE STUDY OF LOW-VAPOR-PRESSURE 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 Daniel Robert Haylett Marh 2011

2 0BABSTRACT This thesis desribes a new faility and method of experimentation, whih an be used to study the ombustion hemistry of low-volatility fuels in the gas phase. Two main goals are desribed: first, the development of the aerosol shok tube and proedures; and seond, a demonstration of its apabilities. There is a lak of high-quality, aurate hemial kinetis data for the oxidation of large hydroarbons, whih are important for modeling diesel, roket, or jet engines among other ombustion systems. While onventional shok tubes are very effetive reator vessels for low-moleular-weight gaseous fuels (n-alkanes up to five arbon atoms), larger fuel moleules exist as low-volatility liquids/solids, and the vaporpressures of these fuels are not large enough for high or even moderate fuel loadings. Heating the shok tube has extended the use of shok tubes to arbon numbers of 10 to 12, but beyond that, the high temperatures prior to the shok initiation an deompose the fuel, and (for fuel mixtures like diesel) an ause frational distillation. The question is then: how an we study low-vapor-pressure fuels in a shok tube? The solution presented here, whih avoids the problems assoiated with heating, is alled the aerosol shok tube. In the aerosol shok tube, the fuel is injeted as an aerosol of miron-size droplets. Then a series of shok waves first evaporate the fuel and subsequently raise the resultant purely gas-phase mixture to ombustion-relevant iv

3 temperatures. With proper seletion of the shok strength and timing, this proess effetively deouples the mass and heat transfer proesses assoiated with evaporation from the hemial mehanism of ombustion. This enables the study of extremely lowvolatility fuels, never before studied in a purely gas-phase form in a shok tube. The first appliation of this new faility was to measure the ignition delay time for many previously inaessible fuels in the gas-phase. In this thesis, we have measured ignition delay times for the pure surrogate fuel omponents n-deane, n-dodeane, n- hexadeane, and methyl deanoate as well as for multi-omponent fuels suh as JP-7 and multiple different blends of diesel fuel. Taken over a range of onditions, these measurements provide sensitive validation targets for their respetive hemial mehanisms. These data showed agreement with past heated shok tube experiments for fuels in whih premature fuel deomposition is not an issue (n-deane and low onentration n-dodeane). However, when omparing heated and aerosol shok tube ignition delay times for fuels that require signifiant heating, like n-hexadeane, the existing heated shok tube data demonstrated evidene of premature deomposition. The seond appliation to the study of hemial kinetis was to measure the onentration of important speies during the deomposition and oxidation of selet lowvapor-pressure fuels. These speies time-histories provide muh more information for kineti mehanism refinement. Experiments were performed to measure the important OH radial and the stable intermediate C 2 H 4 for both n-hexadeane and diesel. The number of important low-vapor-pressure fuels that require high-quality validation targets is large, and our new method for providing this data has proven very effetive. This work enables the development of the next generation of aurate hemial mehanisms and will be essential to their suess. v

4 1BACKNOWLEDGEMENTS First of all I would like to thank my advisor Ron Hanson for his ontinued inspiration and motivation for suess in my work. He was helpful not only in the advie he gave, but also in the example of hard work and determination that he set. I would also like Dave Davidson for his ontinued support and enthusiasm. He is an integral part of our lab. I am very thankful to have had his help in this work and in disussions about life. The Hanson group is like a team, and along with the very helpful post-dos Petros Lappas and Guillaume Pilla, I have to thank some of the senior students: Tom Hanson, Dave Rothamer, Adam Klingbeil, Ethan Barbour, Greg Rieker, Aamir Farooq, among others for their help and support. All of these people have not only been helpful in my researh, but have also been friends outside of the lab. I would also like to give thanks to Megan MaDonald, Jason Porter, Zekai Hong, Jon Yoo, Matt Campbell, Brian Lam, and all my other friends and oworkers. My parents Jim and Debbie Haylett have always been there supporting me and enouraging my sientifi pursuits, as have my brother Jim Haylett and sister Andrea Haylett. Most of all, I would not have made it through this proess without the love and support of my wife Gemma, whom I met during this work. Her willingness to share in my goals and help in keeping me grounded was a key part to making this work suessful. vi

5 2BCONTENTS ABSTRACT ACKNOWLEDGEMENTS CONTENTS LIST OF FIGURES IV VI VII XI LIST OF TABLES XIX CHAPTER 1: INTRODUCTION Distillate Fuels Surrogate Fuels Shok Tubes Aerosol shok tube Conlusions...13 CHAPTER 2: COMBUSTION STUDIES IN THE AEROSOL SHOCK TUBE Aerosol Generation Loading the Aerosol into the Shok Tube Measuring Shok Speed Fuel Measurement In-Situ Measurement of Droplets Gas-Phase Laser Absorption Fuel Spetrosopy FTIR/Heated Cell Measurements Shok tube measurements of high temperature ross Setion Calulation of Equilibrium Shok Jump Conditions Validation of AEROFROSH...31 vii

6 2.6 Appliations of Measurements Behind Refleted Shok Waves Conlusions...36 CHAPTER 3: UNIFORMITY IN THE AEROSOL SHOCK TUBE Introdution Aerosol Shok Tube Generation II Experimental Apparatus Calulation of Loading Pressures Results Nebulizer Performane Mixing Tank Loading Uniformity Shok Tube Loading Uniformity Shok Wave Experiments Fuel Loading Uniformity Ignition Delay Time Measurements Conlusions...59 CHAPTER 4: APPLICATION 1 - IGNITION DELAY TIMES Introdution Low-Vapor-Pressure Fuels n-deane n-dodeane n-hexadeane Diesel (DF-2) Methyl Deanoate Experimental Setup Results n-deane n-dodeane n-hexadeane Methyl Deanoate Diesel...75 viii

7 4.5 Conlusions...79 CHAPTER 5: APPLICATION 2 - SPECIES TIME-HISTORIES Introdution Experimental Setup Aerosol Shok Tube Laser Absorption Results and Disussion n-hexadeane Oxidation Diesel Oxidation Conlusions...92 CHAPTER 6: CONCLUSIONS Faility and Method Ignition Delay Times Speies Time-Histories Future Work Negative Temperature Coeffiient Biofuels...96 ix

8 APPENDIX A 97 A.1 Liquid films...97 A.2 Window Design...99 A.3 Operational Conditions...99 APPENDIX B 101 B.1 Shok aerosol interations B.2 Gas Phase Diffusion B.3 Lab Time and Partile Time APPENDIX C 108 C.1 Behind the Refleted Shok C.2 Test Time C.3 Contat Surfae Refletion C.4 Constant U, V Operation APPENDIX D 112 D.1 Detailed Calulation of Initial Pressures APPENDIX E 115 E.1 Ignition Delay Time Data APPENDIX F 120 F.1 AEROFROSH Code REFERENCES 143 x

9 3BLIST OF FIGURES Figure 1.1: Distillation urves for ommon distillate fuels as well as a soybean derived bio-diesel Figure 1.2: Past ignition delay time measurements normalized to 6 atm and an equivalene ratio of 0.5. Many tehniques and failities were used to measure these ignition delay times Figure 1.3: Chain length distribution of n-alkanes present in two different bathes of diesel fuel [12] Figure 1.4: Fuel surrogate vapor pressures. Heated shok tube temperature limits partial pressure of fuel that an be added to mixtures Figure 1.5a-: a. Shok tube prior to shok. Driver is filled till diaphragm breaks. b. Inident shok travels into driven setion followed by ontat surfae. Expansion wave travels into driver.. The inident shok is refleted at the endwall and forms a refleted shok whih travels in opposite diretion reating ombustion-relevant onditions Figure 1.6: Aerosol shok tube prior to initiation of shok. Liquid or solid partiles are loaded uniformly in the driven setion to experiment on low-volatility fuels Figure 2.1: Cross-setion shemati of nebulizer in operation. The nebulizer uses a vibrating piezoeletri transduer to pinh off droplets and reate a fog Figure 2.2: 12-disk nebulizer bank in operation. Fuel is n-dodeane Figure 2.3: Experimental setup used to measure droplet size distributions for various fuels Malvern Sprayte TM (Model RTS5214) Figure 2.4: Droplet size distributions for water and representative fuels used in this study Figure 2.5: Shemati of first-generation aerosol shok tube (AST I) filling method. Aerosol spatial uniformity was insuffiient for high quality ombustion experiments. Chapter 3 disusses an improved design Figure 2.6: Attenuation of shok veloity in a dry argon shok. Attenuation rate is 2.1%/m. Shok onditions: P 1 = 0.26 atm, T 1 = 295 K, P 2 = 1.6 atm, T 2 = 715 K xi

10 Figure 2.7: Attenuation of shok veloity in a n-dodeane aerosol. Attenuation rate is 3.6%/m. Shok onditions: P 1 = 0.18 atm, T 1 = 294 K, P 2 = 1.4 atm, T 2 = 695 K, X f = 0.63%, X O2 = 21%, balane Ar, Droplet loading by volume: 10.0 ppmv (volume of liquid per volume of gas) Figure 2.8: Shemati of Mie-sattering diagnosti for measurement of liquid volume onentration Figure 2.9: Shemati of gas phase laser absorption measurements Figure 2.10: Experimental setup for measuring ross setions in a heated ell with an FTIR Figure 2.11: Methyl deanoate spetra measured at various temperatures in a heated ell using a FTIR Figure 2.12: Summary of ross setion measurements at 3.39 µm for various fuels measured using the heated ell and FTIR Figure 2.13: Cross setion measurements of n-dodeane at 3.39 µm extended to high temperatures using the aerosol shok tube Figure 2.14: Cross setion measurements of JP-7 at 3.39 µm extended to high temperatures using the aerosol shok tube Figure 2.15: Example of fuel measurement behind an inident shok wave. The red trae shows droplet sattering and absorption. The blak trae shows only droplet sattering. Arrival of the inident shok ours at 0 µs and arrival of the refleted shok ours at 500 µs. Conditions: 21% O 2 in Ar with X f (n-dodeane mole fration) = , T 2 = 617 K, P 2 = 1.11 atm, T 5 = 993 K, P 5 = 4.23 atm. Measurement loation is 10 m from endwall Figure 2.16: Comparison of AEROFROSH alulations to those made by Guha in 1992 [18] and Marble made in 1968 [19] for a shok traveling in pure water of varying initial quality Figure 2.17: Comparison of measured pressure (2 m from the endwall) with those predited by AEROFROSH Figure 2.18: Comparison of measured temperature using a two-wavelength fuel diagnosti with alulated temperature xii

11 Figure 2.19: Comparison of measured temperature using a two-line CO 2 tehnique with alulated temperature Figure 2.20a,b: a. Plots of pressure versus time measured 2 m from the endwall in AST I for mixtures of n-dodeane / 21% O 2 / Ar at various temperatures = 1.0, P = 6.0 atm. b. Summary of ignition delay times plotted on an Arrhenius plot Figure 3.1a-d: (a) Diagram illustrating omponents of AST II and initial pressures set for operation. (b) Aerosol mixture being prepared in the mixing plenum. () Endwall gate valve and ball valve opened and aerosol mixture expanded into the test setion. (d) Endwall gate valve and ball valve losed and the driven gate valve opened; the shok tube is ready for initiation of the shok wave Figure 3.2: Shemati of filling proess for seond generation aerosol shok tube. Initial pressures in the test setion (V TS ) and mixing tank (V MT ) are both P A and are separated from the dump tank (V B ) by a ball valve, whih is at a lower pressure P B. When the ball valve is opened the pressures equalize to P 1. At this point the expansion is suffiient for the gas from the mixing tank to fill the test setion Figure 3.3: The optial setup on the mixing plenum to measure the uniformity and the loading Figure 3.4: Uniformity in tank: water aerosol. Two lasers at 10 and 40 m from the endwall. The disrepany between the two measurements indiate the magnitude of the non-uniformities present in the mixture at a partiular time. The oeffiient of variation (COV=StDev/Ave) is plotted as a funtion of time Figure 3.5: Filling the tank to different onentrations: water aerosol. Various urves represent different lengths of time that the nebulizer and mixing fan was turned on. This shows the effets of leaving the nebulizer running for different amounts of time and the range of ahievable onentrations (1-100 ppmv) Figure 3.6: Optial setup for measuring non-uniformities in AST II method of filling. 50 Figure 3.7: Plot of laser extintion measurements at three loations along the tube and the resulting oeffiient of variation: water aerosol. AST II generates nonuniformities muh smaller than AST I, in this ase 2% Figure 3.8a-: These plots show measurements of aerosol onentration at three loations while filling the shok tube using the AST II filling method (a) 50 slpm fill xiii

12 rate (b) 4.0 slpm fill rate () 1.0 slpm fill rate. The flow rate was varied by using various diameter orifies Figure 3.9: The relationship between the resulting non-uniformity to the flow rate with whih the tube was filled. The flow rate was varied by using various diameter orifies. The optimal flow rate range is between 2 and 5 slpm (gas veloity ~45m/s) Figure 3.10: Top View: Diagnostis used in aerosol tube experiments. Pressure measurements are used to measure shok speed and ignition delay time. Laser light at 650 nm an be used to measure liquid aerosol onentration. Laser light at 3.39 µm is used to measure gas phase onentration after the fuel has evaporated. Side View: Inident shok propagates into aerosol evaporating and mixing fuel and oxidizer. This region is alled region Figure 3.11a-: Three shok experiments highlighting region 2 post-evaporation uniformity for a range of fuel loadings all with n-dodeane in 21%O2/Ar. (a) Conditions: X f (n-dodeane mole fration) = , T 2 = 617 K, P 2 = 1.11 atm, T 5 = 993 K, P 5 = 4.23 atm (b) Conditions: X f = , T 2 = 584 K, P 2 = 1.64 atm, T 5 = 921 K, P 5 = 6.25 atm () Conditions: X f = , T 2 = 534 K, P 2 = 1.62 atm, T 5 = 838 K, P 5 = 6.26 atm Figure 3.12: Ignition delay times for AST I [21] and AST II. AST II data produes signifiantly redued satter and slightly lower mean values than AST I data. Pressure and Φ (equivalene ratio) were normalized to 5.0 atm and Φ=0.5 using P and Φ dependenes, respetively Figure 4.1: Shemati of the aerosol shok tube with pressure and laser diagnostis. The pressure sensors are used for shok speed measurement and ignition delay time determination, the mid-ir HeNe laser is used for absorption-based fuel measurements, the visible laser diode is used for droplet sattering measurements, and the emission measurement is used to measure ignition delay time Figure 4.2: Example of an ignition delay time measurement. This example was done with a DF-2 (CI 43) / 21% O 2 / argon mixture with = 0.48, T 5 =1197K, and P 5 =7.21 atm. The diagnostis in the upper frame (a) are loated 3m from the endwall, and in xiv

13 the lower frame (b) are 5m from the end wall: (1) Mie sattering extintion, (2) pressure, (3) CH* emission, (4) fuel absorption, and (5) Mie sattering extintion Figure 4.3: n-deane/air ignition delay times. Data onditions ranging over P= atm and = are saled to P=5 atm, phi=1.0 using the Olhanski and Burat orrelation [61] (blue diamonds: data; blue line: orrelation; blak lines: best fit to Shen et al. [48]) Figure 4.4: n-dodeane/air and n-dodeane/21% O 2 /Ar ignition delay times for fuellean mixtures ( =0.5) at various pressures Figure 4.5: Variation of n-dodeane/21% O 2 /Ar ignition delay time with equivalene ratio saled to 6.0 atm ( atm) Figure 4.6: Equivalene ratio dependene of the ignition delay times for n-dodeane at a) 1250K and b) 1000K, both at P=6.0 atm Figure 4.7: (a) Comparison of n-hexadeane ignition delay times at 4 atm and equivalene ratio of 1.0 for various oxygen onentrations. (b) Ignition delay time variation with pressure for stoihiometri n-hexadeane / 4% O 2 / Ar mixtures. () Ignition delay time variation with equivalene ratio at 4 atm for stoihiometri n- hexadeane / 4% O 2 / Ar mixtures Figure 4.8: Ignition delay times of methyl deanoate / 21% O 2 / Ar mixtures at = 0.1 and P=8atm and the ignition delay times from the LLNL mehanism by Herbinet et al. [46] Figure 4.9: Diesel ignition delay times for three different diesel fuels ompared with simulations using the LLNL mehanisms for various surrogate mixtures. Mixtures have an oxygen onentration of 21% with P=6atm and =0.5. Colored, square symbols indiate data taken with aerosol shok tube; blak irles represent data taken in a heated shok tube by Penyazkov et al. [55] Figure 4.10: Variation of ignition delay time with pressure ( =0.5) and equivalene ratio (P=6atm) for mixtures onsisting of DF-2 (CI 43) / 21% O 2 / Ar Figure 4.11: Variation in ignition delay time with aromati ontent of diesel fuel. Oxygen onentrations of 4% were used with Ar as the diluent. P=6atm and =0.5. The two diesel fuel blend measurements were ompared to simulations of mixtures of n-hexadeane and iso-etane (CN 42, 46) using LLNL mehanisms [43, 44] xv

14 Figure 5.1: Shemati of aerosol shok tube (AST II) setup. The different regions in the tube are (from left to right): A - driver gas, B - inident-shok heated driven gas with no fuel, C - inident-shok heated test mixture with fuel, and D - refleted-shok heated test mixture Figure 5.2: Diesel OH on-line and off-line laser absorption near nm and the differene absorbane signal attributable solely to OH during diesel oxidation. The pre-ignition plateau orresponds to around 10 ppm of OH. Initial refleted shok onditions: ONLINE: 1198 K, 6.69 atm, 1228 ppm diesel; OFFLINE: 1193 K, 6.60 atm, 1197 ppm diesel both in 4% O 2 / argon Figure 5.3: OH and C 2 H 4 speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: =1.2, 1267 K, 6.54 atm; initial test gas mixture: 497 ppm C 16 H 34, 1% O 2 /argon. The onstant UV simulation is based on a LLNL hexadeane mehanism by Westbrook et al. [43] Figure 5.4: C 2 H 4 speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: =0.8, 1170 K, 4.60 atm, 326 ppm C 16 H 34, 1% O 2 /argon; =1.2, 1267 K, 6.54 atm, 497 ppm C 16 H 34, 1% O 2 /argon; and =1.2, 1333 K, 6.77 atm; 493 ppm C 16 H 34, 1% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] with onstant UV and aounting for faility dp/dt Figure 5.5: OH speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: =0.8, 1170 K, 4.60 atm, 326 ppm C 16 H 34, 1% O 2 /argon; =1.2, 1267 K, 6.54 atm, 497 ppm C 16 H 34, 1% O 2 /argon; and =1.2, 1333 K, 6.77 atm, 493 ppm C 16 H 34, 1% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] with onstant UV and aounting for faility dp/dt Figure 5.6: OH and C 2 H 4 speies time-histories during diesel oxidation. Initial refleted shok wave onditions: =0.54, 1198 K, 6.69 atm; initial test gas mixture: 1228 ppm diesel, 4% O 2 /argon. Constant UV simulations for single-omponent surrogate models shown are: n-heptane and n-dodeane whih use the JetSurf mehanism and n-hexadeane whih uses the LLNL mehanism Figure 5.7: C 2 H 4 speies time-histories during diesel oxidation. Initial refleted shok wave onditions: =0.8, 1119 K, 4.63 atm, 1761 ppm diesel, 4% O 2 /argon; =0.5, 1198 K, 6.69 atm, 1228 ppm diesel, 4% O 2 /argon; and = K, 6.33 atm, 1380 xvi

15 ppm diesel, 4% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] for n-hexadeane as the fuel with onstant UV and aounting for faility dp/dt Figure 5.8: OH speies time-histories during diesel oxidation. Initial refleted shok wave onditions: =0.8, 1119 K, 4.63 atm, 1761 ppm diesel, 4% O 2 /argon; =0.5, 1198 K, 6.69 atm, 1228 ppm diesel, 4% O 2 /argon; and = K, 6.33 atm, 1380 ppm diesel, 4% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] for n-hexadeane as the fuel with onstant UV and aounting for faility dp/dt Figure A.1: Examples of shok veloity attenuation for various window heating onfigurations. Note that the old heater raised the temperature in region 1 aelerating the shok near the endwall Figure A.2: New heated window mount prior to installing on shok tube. Copper tubes are wrapped with ring heaters whih ondut heat to windows. The mount is ooled by flowing ool water through inside of mount Figure B.1: X-t diagram for shok aerosol interation. Partile breakup ours fastest with partile aeleration ourring more slowly. Evaporation and diffusion our most slowly after whih the flow is in equilibrium. The measurement loation is seleted suh that all these proesses have ourred to a suffiient extent upon arrival of the refleted shok Figure B.2: Evolution of gas phase fuel onentration. Assuming spherial symmetry and that the droplet instantaneously evaporates. In this ase, it takes only 15 µs for the fuel onentration to be 95% uniform Figure B.3: Time to ahieve 95% uniformity with varying volume fration of liquid for monodisperse size distribution of various diameters Figure C.1: Piture of aerosol shok tube driver setion and additional segments for extending driver length (Photo Courtesy of Matt Campbell) Figure C.2: Pressure trae from pressure transduer loated at 2 m from the endwall on a test using a tailored driver mixture. At 0 ms the inident shok arrives followed losely by the refleted shok. In region 5 the pressure slowly inreases by some non-ideal faility effets. At 6 ms the refletion off the ontat surfae arrives but with no signifiant pressure jump. The pressure trae is reasonably level until the arrival of the expansion wave at 11 ms xvii

16 Figure C.3: Figure 4.9 of ignition delay time plot vs. inverse of temperature showing the effet of modeling as onstant UV or with a 0.2 atm/ms dp/dt. The effet beomes apparent at test times above 10 ms Figure D.1: Simplified shemati of the expansion proess showing (a) the initial state and (b) the final state Figure D.2: Measurements of final pressure (P 1 ) after expansion of gas/aerosol mixture into dump tank. Initial pressure in the aerosol mixing tank and test setion (P A ) was around 700 torr and the initial pressure in the dump tank (P B ) was varied. Also shown are various models used to predit the expansion behavior. The hybrid model that assumes isothermal expansion and isentropi ompression seems to predit the final pressure quite well Figure D.3: Plot indiating how to set initial pressures in the mixing tank/test setion and the dump tank. For example, to ahieve a equilibrium pressure (P 1 ) of 100 torr and expansion fator of 2, the mixing plenum and test setion should be set initially to 185 torr (P A ) and the dump tank should be set initially to 62 torr (P B ) xviii

17 4BLIST OF TABLES Table 2.1: Tabulated results from measurement of temperature using the CO 2 diagnosti. Results show good agreement between measured temperature and predited temperatures using AEROFROSH Table E.1: Table shows ignition delay time data with assoiated onditions. See hapter 4 for analysis xix

18 Chapter 1: 5BINTRODUCTION In the United States, the ombustion of fossil fuels aounts for over 80% of our energy usage at a ost of nearly $1 trillion per year [1]. Thus any marginal improvement in ombustion effiieny an be translated into sizable savings eonomially as well as environmentally. To ahieve this, we strive to understand the proess of ombustion at its most fundamental level. When modeling the ombustion in engines, fluid transport and heat transfer are important. But ombustion is ultimately ontrolled by hemistry, and therefore a hemial mehanism needs to be well-founded for a model or CFD pakage to be suessful. A hemial mehanism is a road map that eluidates the exat hemial pathways by whih fuel and oxidizer are transformed into produts. These pathways begin with the fuel moleules breaking apart. Muh of the fuel used is in the form of very large hydroarbon moleules; for example, gasoline, jet-a, diesel, and bio-diesel have average arbon numbers of roughly 7, 12, 16, and 18 (per moleule), respetively. The large moleular size makes the modeling and experimentation of hemial kineti phenomena diffiult: modeling, beause the number of speies and reations formed during oxidation 1

19 inreases; and experimentation, beause these large moleules have low-volatilities and deoupling the evaporation proess from the ombustion proess is very diffiult BDISTILLATE FUELS Low-vapor pressure fuels have been and still are very widely used in our soiety. Their low volatility at atmospheri onditions make these fuels attrative due to a lower risk of unintended explosion, however, it also makes them diffiult to work with experimentally. Diesel fuel is one of the most widely used petroleum-based fuels used today. It is ommon for large, high-output engines to operate on diesel fuels. Commerial truks and SUV s, large industrial equipment suh as bulldozers, front-end loaders, dump truks, boats, loomotives, submarines, eletri-power generators, et., all use diesel fuel. This is beause the diesel yle is more fuel effiient than the Otto yle; however, some may argue that diesel engines are more harmful to the environment [2]. More reently lean diesel fuel has beome more popular, and ompanies like Meredes-Benz, Honda, Nissan, and Toyota are starting to produe passenger vehiles that run on diesel fuel in the US. Lower equivalene ratios, tougher standards on fuel omposition, and better atalyti onverters have all greatly improved the environmental aspets of using diesel fuel. Despite these advanes, we still do not have an aurate detailed hemial mehanism that desribes the ombustion hemistry that takes plae inside a diesel engine. Diesel fuel is a type of distillate fuel reated by refining rude oil. Crude oil is a mixture of thousands of hydroarbon moleules. The refining and raking proess begins by raising the rude oil temperature. As the temperature rises some omponents start to hange phase from a liquid to a gas. Eah omponent has a different response to this temperature inrease so the hemial omposition of the liquid hanges with time. The gas-phase moleules are ondensed and then olleted in a separate vessel. Depending on what temperature is hosen to start and end this proess, rude oil an be separated into many different uts and used for many different purposes. There are 2

20 ertain standards whih define a partiular distillate, but nonethelesss there is variability in every bath. Some examples of distillation urves for popular fuelss are shown in Figure 1.1. The data shows the volume of liquid reovered by the ondenser at a given temperature. Gasoline is a lower temperature ut between C, jet fuel is a slightly less volatile fuel whih is ut between C, and diesel is one of the higher temperature fuels at C. Bio-diesel, also shownn here, is not derived from rude oil, but is manufatured from plant matter. A typial soybean-derived but these are the main types of bio-diesel is evaporated between C. There are many others, fuels thatt are distilledd from rude oil. Figure 1.1: Distillation urves for ommon distillate fuels as well ass a soybean derived bio-diesel. The high-temperature evaporation window of diesel is what makes it ideal to use in the diesel yle and also what makes it very diffiult to study. Combustion takes plae at temperatures between K, and at these onditions all of the omponents of diesel fuel are in the gas phase in equilibrium. Therefore in order to do a purely gas- before phase ombustion hemistry experiment, the fuel needs to undergo a phase hange the experiment. To do this we must heat the fuel up to at least 400 C to ensure it is ompletely in the gas phase. Although rapid oxidation does not our at these temperatures, slow deompositio on and pyrolysis an our (and probably do our in the distillation proess). 3

21 Many of the previous studies of diesel fuel ombustion suffer from the effets of evaporation, whih ompliate the interpretationn of the measurement. Some researhers attempt to avoid this by pre-heating the fuel. Heating the fuel too muh will prematurely deompose the fuel, whereas failing to raise the temperature high enough will frationally distill the fuel, thereby altering the omposition of the fuel in the gas phase. An example of an experimental measurement that iss intimately onneted to the overall hemistry is the ignition delay time. This is the time that it takes for reatants at a ertain pressure and temperature to auto-ignite. Thesee measurements an provide ruial targets for the validation of kineti models; however, for low-vapor-pressure fuels few studies of this type exist. The data that do exist are quite sattered. Representative data are shown in Figure 1.2. Figure 1.2: Past ignition delay time measurements normalized to 6 atm and an equivalene ratio of 0.5. Many tehniques and failities were used to measure these ignition delay times. 4

22 The omparative study of diesel fuel ignition delay time is ompliated by the different types of experiments performed over the years, and the simple lak of data on neat diesel itself. The studies that have been published an be divided into two ategories: those done with spray injetors in shok tubes, rapid ompression mahines (RCM) and engines; and those done with flow tube reators. Kobori et al. [3] studied diesel surrogate sprays in a RCM (P= atm, =1). Clothier et al. [4] studied diesel ignition delay times in a single-ylinder gasoline engine modified for diesel operation (P=16 atm, =1). Boiko et al. [5] investigated the ignition of droplets (typially 2 mm diameter) behind refleted shok waves (P=23 atm, =1). Tsuboi and Kurihara [6] studied sooting of a light-fuel-oil spray injeted into a shok tube (P=10 atm, =1). Mellor et al. [7] desribe the fuel-spray shok tube work of Hurn and Hughes [8] for partially refined diesel fuels as well as other early work (P=35 atm, =1). Tahina [9] studied diesel ignition in a flow reator. Spadaini and Tevelde [10] (P=10-30 atm, =0.3-1) and Tevelde and Spadaini [11] (P=3-5 atm, =0.2-1) advaned this method and used a speialized flow reator with fully evaporated flows to study diesel ignition times at elevated pressures. The range of these works is shown in Figure 1.2; the data have been normalized to 6 atm and an equivalene ratio = 0.5 using the orrelation of Spadaini and Tevelde [10]. The ignition delay times generated using spray injetors in shok tubes and a RCM are longer than the flow reator data, beause they inlude the additional time needed to evaporate and mix/diffuse the fuel/air mixture, as well as for ignition to our. These diesel injetor spray studies provide information about the relative ignition properties of different fuels, but analysis of the data is ompliated beause the hemistry of ignition of these sprays is strongly onvolved with the evaporation and flame droplet proesses. As a onsequene of the large size of the single droplets, the strongly varying spatial distribution, and highly non-uniform size distribution of the diesel injetor sprays, details of the shok-spray interation and the hemistry of these experiments are not easily quantified. On the other hand, the flow tube reator data, represented by 5

23 Spadaini and Tevelde [10,11] overs the lower temperature regime and attempts to remove the effets of evaporation by preheating the fuel. No shok tube diesel fuel ignition delay data appear to exist for pre-evaporated fuel mixtures BSUR RROGATE FUELS Along with the study of diesel fuel itself, it is important to also study the individual omponents of diesel fuel. Most ruial are those omponents thatt an appropriately represent the harateristis of diesel fuel. It is impossible to model every omponent of diesel fuel; therefore, ommonlyy aepted representative surrogates are often used to model the ombustion behavior in useful devies. Distillate fuels an be thoughtt of as beingg omprised of six different ategories of iso-alkenes, and hydroarbons: normal-alkanes, iso-alkanes, ylo-alkanes, alkenes, aromatis. A good surrogate mixture inludes representative fuel moleules from eah of these ategories. In the ategory of n-alkanes, the most prevalent moleule is n- hexadeane or etane, whose moleular formula is C 16 H 34. A plot of the weight fration versus arbon number for the n-alkanes reveals a bell urve entered near C16. This is shown for two different bathes of diesel fuel in Figure 1.3 (data taken from Farrell et al. [12]). Figure 1.3: Chain length distribution of n-alkanes present in two different bathes of diesel fuel [12]. 6

24 If we look at the other fuel ategories, we also find high arbon number omponents. For example in the iso-alkane ategory, iso-etane (2,2,4,4,6,8,8 heptamethylnonane, an isomer of n-hexadeane) is also an important fuel for understanding diesel ombustion. In fat, a mixture of just these two omponents (etane and iso-etane), is the most widely used representative mixture for the ignition behavior of diesel fuel. A mixture of etane and iso-etane makes up what is alled the primary referene fuel (PRF) for diesel, whih provides a standard definition for the etane number of a fuel. To measure a partiular fuel s etane number, the fuel s ignition delay time is measured in a Cooperative Fuel Researh (CFR) engine. Then the ignition delay time is mathed with that of a partiular mixture of etane and iso-etane. Cetane has a rapid ignition delay and it is arbitrarily given a etane number (CN) of 100 whereas iso-etane has a slower ignition delay and it has a value of 15 (the definition for the lower bound etane number omes from alpha methyl naphthalene CN=0). The mixture etane number is a weighted average of its omponents etane numbers. The other ategories of diesel representative fuel omponents an be found in Farrell et al. [12]. For aromatis, n-deyl benzene and 1-methyl napthalene were reommended. Others have suggested the use of butylylohexane and dealin in the ylo-alkane ategory. Another important omponent is tetralin. There have been very few ombustion studies of these representative fuel omponents. This again is primarily due to the fat that all of these omponents, like diesel fuel, have very low vapor pressures. For n-hexadeane there have been some rapid ompression mahine studies [13], but there has been only one shok tube study [14]. On the other hand, n-alkanes with arbon numbers one through four (i.e. methane (C1) through n-butane (C4)) are gas phase at atmospheri onditions, so these are relatively easy to study. n-pentane (C5) through n-heptane (C7) are liquids at atmospheri onditions, but have high vapor pressures; therefore, it is still possible to fill a hamber with gaseous fuel up to the saturation vapor pressure at the temperature of the experiment (300K). 7

25 When the arbon number is raised even further the vapor pressure drops and this starts to limit the fuel loading to a point where it makes ertain high fuel loading onditions inaessible in shok tubes for fuels as small as n-otane (C8). One ommon method of dealing with this issue is to heat the entire experimental apparatus in order to raise the saturation vapor pressure of the fuel and avoid ondensation. This is very effetive and greatly extends the test spae of fuels up to n-dodeane (C12), but beomes prohibitive at higher arbon numbers, beause the fuel starts to deompose by raking and pyrolysiss when exposed to elevated temperatures for extended periods of time (e.g. the time it takes to make a mixture of fuel and oxidizer). This time varies based on the mixing tank used, but is typially on the order off hours to reate a homogeneous mixture. Vapor pressures of seleted speies as a funtionn of temperature are shown in Figure 1.4. For a typial mixture, partial pressures of 10 to 100 torr of fuel are generally needed. Reently Penyazkov et. al. [14] has used a heated shok tube to study the ignition behavior of n-hexadean ne (C16), and Ristori ett al. [13] have studied ignition times of hexadeane (C16) in a rapid ompression mahine. Based on this work, n-hexadeane appears to be the limit of what is possible with heating. The aessible fuel loading onditions for a very heavy n-alkane, e.g. n-eiosane (C20), would be severely limited using a heated experiment. The need for a better method of testing low-vapor-pressure fuels is apparent for surrogates as well as for diesel. Figure 1.4: Fuel surrogate vapor pressures. that an be added to mixtures. Heated shok tube temperature limits partial pressure of fuel 8

26 1.3 20BSHOCK TUBES Shok tubes have been in use for more than a entury and have been prolifi in studies ranging from high energy physis and wind tunnels, to hemial kinetis. The study of hemial kinetis in ombustion systems is well-suited for shok tubes. The shok tube provides a method to raise the temperature and pressure to a well-ontrolled value almost instantaneously. The gas is also stagnant and adiabati providing experimental onditions behind the refleted shok that an be modeled as onstant internal energy and onstant volume (U, V) when there is negligible hemial energy released. As shown in Figure 1.5a, a shok tube is basially a long losed tube, whih is hermetially separated into two parts: a driver setion and a driven setion. At the partition between these two setions there is a diaphragm that is designed to burst at a speifi pressure differene between the two setions. The driven setion is filled with the test gas mixture, and the experiment is initiated by filling the driver with helium (and sometimes a mixture of helium and nitrogen) until the diaphragm bursts. At this point the sharp pressure differene at the rupture reates a shok wave (inident shok) whih travels into the driven setion, as well as an expansion fan that travels into the driver setion (Figure 1.5b). The driven setion is made long enough suh that the shok wave an beome stable and flat. Behind the inident shok the temperature of the driven gas is elevated within a few ollisional pathlengths and is also aelerated toward the endwall (this is alled region 2). When the inident shok reahes the end of the driven setion it hits the endwall ap. This reflets the shok and a new refleted shok wave is reated and travels in the opposite diretion (Figure 1.5). Beause the endwall stops the flow, the refleted shok stagnates the gas headed toward the endwall. This deeleration aused by the refleted shok further heats and ompresses the gas (this is alled region 5, see Appendix B for an X-t diagram). Near the endwall the gas has stagnated at elevated pressures and temperatures with a well-defined time zero, where it an be easily 9

27 studied with various diagnostis. These onditions are only limited by the eventual return to equilibrium brought on by suessive refletions of the expansion wave. Muh insight has been gained in the studyy of ombustion using shok tubes. One an measuree individual elementary reation rates, ignition delay times, the high temperature properties of ombustion relevant moleules, and muh more. Gaydon and Hurle have written a seminal text on the operation and dynamis of shok tubes used for these purposes [15]. A few tens of researh groups around the world operate shok tube laboratories to study ombustion kinetis. a b Figure 1.5a-: a. Shok tube prior to shok. Driver is filled till diaphragm breaks. b. Inident shok travels into driven setion followed by ontat surfae. Expansion wave travels into driver.. The inident shok is refleted at the endwall and forms a refleted shok whih travels in opposite diretion reating ombustion-relevant onditions. 10

28 1.4 21BAEROSOL SHOCK TUBE The benefits of shok tube reators an be utilized if we apply them to the study of low-vapor-pressure fuels. In this work, we have developed a method and faility for applying the shok tube tehnique to low-vapor-pressure fuels, whih is alled the aerosol shok tube. In this faility, the fuel is loaded in the form of an aerosol and the inident shok evaporates the fuel. The aerosol shok tube onept is an extension of the onventional shok tube tehnique. Reall that the heated shok tube tehnique would involve heating the shok tube and mixing assembly to allow a higher fuel vapor pressure to exist. The problem with this method appears prior to the shok initiation in the shok tube or mixing tank when the fuel temperature is too high for too long. The fuel will begin to deompose prior to the ombustion experiment during the mixing and filling of the shok tube. The aerosol shok tube method avoids this by reduing the amount of time that the fuel is at elevated temperature prior to the ombustion experiment from hours to frations of a milliseond. We an redue the heating time by taking advantage of the inherent gas-dynamis that our in a onventional shok tube. In shok tubes, the gas mixture being studied undergoes two subsequent shok ompressions, first an inident shok, whih approximately doubles the temperature (~600K) in urrent experiments and next the refleted shok whih again raises the temperature by another fator of two (~1200K). This intermediate step (region 2) an be taken advantage of in the ase of low-vaporpressure fuels. The temperatures in this region (500K-800K) an evaporate almost any distillate liquid fuel that is in use today (in approximately 100 µs, See Appendix B for details about shok-aerosol interations). Then the refleted shok will bring the purely gas-phase mixture to ombustion-relevant temperatures where the hemistry an be observed. The diffiulty beomes how we load the shok tube with liquid fuel prior to the initiation of the inident shok. The amount of gaseous fuel needed in the typial 11

29 ombustion experiment is small (~1%) ompared to the total amount of gas required. Furthermore, in liquid form, the density is about a fator of 1000 times higher. Thus, only tens of ppm of liquid fuel volume fration is typially needed. This onentration an be ahieved by aerosolizing the liquid fuel using ultrasoni nebulizers and then flowing it into the shok tube (Figure 1.6). One the aerosol is in the shok tube, the shok is initiated. The inident shok propagates as it would normally, but at a slightly lower speed due to the fat thatt the heat of vaporizationn of the droplets ats as an enthalpy sink behind the shok wave. When a droplet is shoked it is quikly evaporated and diffusion rapidly levels out the spikes in onentration (Appendix B). This leaves behind a ompletely uniform, gas-phase the temperature to the point where ombustion an take plae in the test mixture of ombustible gases. Then, the refleted shok further raisess time available. Figure 1.6: Aerosol shok tube prior to initiation of shok. Liquid or solid partiles are loaded uniformly in the driven setion to experiment on low-volatility fuels. This ideal operational view brings up some obvious questions of pratiality and implementation. For example, how do we ensure uniformity in the initial distribution of aerosol? This question was one of the hardest obstales to overome in order to suessfully implement this method. Any spatial non-uniformity in the initial aerosol would greatly affet the uniformity of the final mixture behind the refleted shok. In this thesis, I will first introdue the method of filling that was previously used when the aerosol shok tube was first envisioned as a faility for studying the interation of aerosols with inident shok waves. We all this the first-generation aerosol shok tube (AST I). Then I will desribe the testing andd methodology involved in the seond- 12

30 generation aerosol shok tube (AST II). The improvements made result in improved spatial uniformity of the aerosol in the initial mixture BCONCLUSIONS There is a definite need to expand urrent modeling efforts to larger fuel moleules; however, the experimental hallenge of separating the evaporation proess from the ombustion proess beomes diffiult for these fuels. Little to no experimental data exist for distillate fuels suh as diesel and bio-diesel. The data that does exist for diesel fuel is widely sattered, and a suessful method has not yet emerged to separate the evaporation effets and examine the purely hemial phenomena that our during the oxidation of this fuel. Fuel surrogates have been identified; however, these ompounds also suffer from the hallenges assoiated with low volatility. Consequentially, there have been very few experiments probing the ombustion hemistry of these low-volatility fuels. Some surrogates suh as n-hexadeane have been studied in a heated shok tube, whih uses elevated temperatures to inrease the vapor pressure of the fuel to allow more fuel into the test mixture. However, these studies may be subjet to premature deomposition. We have developed a new method alled the aerosol shok tube, whih takes advantage of shok tube methodology while avoiding the problem of premature deomposition. The shok tube is loaded with an aerosol made up of liquid fuel droplets surrounded by a gaseous oxidizer and diluent. This aerosol is rapidly evaporated behind the inident shok, and then the refleted shok brings it up to ombustion relevant temperatures. This solution presents a unique approah to gas phase experimentation for lowvapor-pressure fuels. For this method to work properly with appliations to ombustion, many tehnial hallenges must be solved. These issues will be disussed in the next two hapters. 13

31 Chapter 2: 6BCOMBUSTION STUDIES IN THE AEROSOL SHOCK TUBE This hapter is devoted to the details of the aerosol shok tube and the hallenges in making it a suessful approah to study the ombustion kinetis of low-vaporpressure fuels. Aerosol shok tube operation begins with the aerosol generation. This proess has been refined and well-haraterized. Then the filling method is desribed. This ditates the spatial uniformity of the aerosol. Filling and uniformity will briefly be introdued for the first-generation filling tehnique, and then will be overed more ompletely for the seond-generation tehnique later in Chapter 3. Essential measurements will then be disussed suh as shok speed measurements, as well as fuel onentration measurements. These measurements are ritial for alulation of the onditions (temperature and pressure) of the ombustion experiment. 14

32 2.1 23BAEROSOL GENERATION At the foundation of this method is the generation of liquid fuel aerosols. There are many different devies available for the generation of aerosols, all of whih produe aerosols with different properties. For example in a diesel engine, the aerosol generator is inorporated in the fuel injetors. These fuel injetors employ a high-pressure pump that fores the fuel through a small orifie, and jetting leads to Rayleigh instabilities, whih break up the jets into droplets. These droplets range from µm in size. The terminal veloity is too fast for the aerosol shok tube, where the aerosol must stay suspended and not be lost to the walls of the shok tube. In order to make smaller droplets, whih take longer to settle, a more advaned method is needed. Many devies operate by the priniple that a piezoeletri material will physially expand and ontrat under the influene of an alternating urrent up to very high frequenies. This effet an be utilized to reate standing waves on the surfae of a liquid that will pinh off into small droplets. If this frequeny is on the order of a megahertz, the resulting droplet diameters are on the order of mirons. This droplet size is more appropriate for the loading of an aerosol shok tube as the settling time is a few minutes. For the aerosol shok tube, we utilize an array of piezoeletri disks submerged in the liquid fuel to aerosolize large quantities of the fuel (See Figure 2.1 for a ut-away of one nebulizer). The devies that ontain the piezoeletri disks are alled ultrasoni nebulizers, and this proess is referred to as nebulizing the fuel. The resultant droplet size distributions of the various fuels are measured using a Malvern Sprayte TM (Model RTS5214) partile sizer. 15

33 Figure 2.1: Cross-setion shemati of nebulizer inn operation. piezoeletri transduer to pinh off droplets and reate a fog. The nebulizer uses a vibrating Thesee banks of nebulizers (Mist Maker TM 12-disk Nebulizer Bank Model DK12NS) are loated outside of the shok tube in a vauum-sealed ompartment. During operation the liquid surfae is somewhat violent, and for optimal aerosol generation the eiling of the ompartment must not ome in ontat with the bulk surfae of the fluid. This requires a eiling height of at least six inhes above the surfae of the liquid. The piezoeletri transduers output so muh energy that large spires or jets are sustained above eah jet. These jets an reah heights of four inhes. These large disturbanes are a result of the loalized foring of the surfae and a spatial low- frequeny response. There also exist standing waves whih represent the spatial high- of the spiress and are responsible for pinhingg off the droplets. The large jets are frequeny response to the vibration. These lookk like tiny lumps or hairs on the surfae ontinually sloshing and breaking-up, and as a result some very large droplets are made (5 mm), but these immediately fall bak into thee liquid pool, and only the miron sized droplets remain. See Figure 2.2 for piture of a nebulizer bank in operation. 16

34 Figure 2.2: 12-disk nebulizer bank in operation. Fuel is n-dodeane. To aurately measure the droplet sizes the aerosol must be made and entrained in a gas flow to transport it to the partile sizer. The experimental setup is shown below in Figure 2..3, and the results of these measurements for water, n-dodeane, and diesel fuel are shown in Figure 2.4. These distributions are all approximately the same. This is due to the fat that droplet size is most intimately related to the nebulizer frequeny, and not as sensitive to the liquid s surfae tension and density. The result is a mass mean droplet diameter around 4 µm with a spread between 1-10 µm. Figure 2.3: Experimental setup used Sprayte TM (Model RTS5214). to measure droplet size distributions for various fuels Malvern 17

35 Figure 2.4: Droplet size distributions for water and representative fuels used in this study BLO TUBE ADING THE AEROSA SOL INTO THE SHOCK Loading the aerosol into the shok tubee homogeneously is a diffiult task; the quality of our ombustion measurements relies heavily on the homogeneity of the mixture. For aerosols, any small spatial non-uniformity translates into a large, gas-phase non-uniformity. I will first disuss the simpler first-generation method of filling used in AST I. The next hapter ontains a desriptionn of the more omplex improvement of this method used in AST II. One these miron-sized partiles are reated, their large settling times and their small Stokes Numbers allow for easy entrainmen in the gas flow. The droplets trak the flow of bathgas very losely. Typially the bathgas is a mixture of an inert gas Ar or N 2 and the oxidizer O 2 in a ratio that is presribed by the ombustion experiment. The bath gas flows into the nebulizer ompartment at orr below the level of the liquid fuel and bubbles to the surfae. From there the gas-aerosol mixture flows through a two inh inner diameter onduit, then throughh poppet valves in the endwall into the shok tube, and ontinues flowing along the entire length of the driven setion to an outlet port in the sidewall of the shok tube near the diaphragm (Figure 2.5). One a steady state flow is 18

36 setup throughout the shok tube, the shok tube is onsidered loaded. To initiate a shok all the valves are losed and the flow is diaphragm breaks and the shok is reated. stopped, then the driver is filled until the Figure 2.5: Shemati of first-generation aerosol shok tube (AST I) filling method. Aerosol spatial uniformity was insuffiient for high quality ombustion experiments. Chapter design. 3 disusses an improved A very important aspet of this design is the poppet valves at the endwall. These valves are speially designed suh that when they are opened they reate turbulene in the flow, and one losed they reate a very flat endwall. The design used was similar to those in an internal ombustion engine for the intake andd output of gases. There are four poppet valves whihh fill the endwall ross-setional area. The poppet valves are opened only slightly to reate jets of high-speed bathgas that are mixed with the low speed gas near the edges of the opening. This mixing reates turbulene, and serves to fill all the voids and orners uniformly with aerosol laden bathgas. Further down the tube the turbulene dampens out and dissipates, and the resultt is plug flow. This plug flow exhibits some unwanted river effets (this is desribed in detail in Tom Hanson s thesis [16]). Essentially as the aerosol makes its way down the tube the aerosol settles and sets up a dense flowing river of aerosol on the bottom of thee shok tube while the top of the shok tube is void of aerosol. 2.3 MEASU 25BM URING SHOCK SPEED With the aerosol in the tube and the valves losed, the shok an be initiated by filling the driver and bursting the diaphragm (See Figure 4.1 for the experimental arrangement of the pressure transduers and Figure 2.6 and Figure 2.7 for representative 19

37 attenuation plots). The shok veloity is arefully monitored using fast pressure sensors that indiate the position and relative timing of the shok. By dividing the distane between transduers by the time interval, we an alulate the veloity. Typially the veloity of the shok dereases as it propagates through the driven setion of the shok tube. This is due to the fat that theree is visous dissipation near the sidewall of the shok tube. The attenuation of the shok is typially linear with about 2%/m attenuationn rate (Figure 2.6). It has been observed that when a shok enounters an aerosol, this further attenuates the veloity (Figure 2.7). The temperature is related to the square of the Mah number, so any error in the veloity measurement will drastially affet the temperature unertainty. Thus having aurate spatially resolved veloity measurements is very important to making quality ombustion hemistry measurements. Figure 2.6: Attenuation of shok veloity in a dry argon shok. Attenuation rate onditions: P 1 = 0.26 atm, T 1 = 295 K, P 2 = 1.6 atm, T 2 = 715 K. is 2.1%/m. Shok 20

38 Figure 2.7: Attenuationn of shok veloity in a n-dodeane aerosol. Attenuationn rate is 3.6% %/m. Shok onditions: P 1 = 0.18 atm, T 1 = 294 K, P 2 = 1.4 atm, T 2 = 695 K, X f = 0.63%,, X O2 = 21%, balane Ar, Droplet loading by volume: 10.0 ppmv (volume of liquid per volume of gas). 2.4 FUEL 26BF MEASUM UREMENT In the aerosol shok tube the onventional method of fuel onentration measurement annot be used. In the onventional method, the fuel onentration is determined manometrially when making the mixture in a separate mixing tank. The mixing tank is equipped with a pressure sensor, and the fuel is added to an empty mixing tank. The pressure is reorded, and then the next omponent is added. The pressure of eah suessive step is reorded, and the mixture mole frations an be alulated using Dalton s Law. The method of measuring pressure to indiate total mole fration does not work in a multi-phase mixture. In the aerosol shok tube to measuree the amount of fuel, we have employed two methods. First we measure the amount of fuel in the liquid phase ( via a size distribution measurement) and add it to the vapor pressure at the measured temperature of the shok tube. After verifying that all the fuel has evaporated behind the inident shok, the fuel absorbane an be measured for the resulting gaseouss mixture. The measurement is typially made near the endwall in the same loation that the ombustion measurements are made. 21

39 BIN-SIT TU MEASUREMENT OF DROPLETS The first method was used extensively in the first-generation aerosol shok tube in the study of the shok-droplet interations, typiallyy in water aerosols [16]. Measurement ts were made using a Mie satteringg light extintion tehnique in a forward- lognormal and varied only slightly in average size and spread, a minimum of three sattering orientation (Figure 2.8). Beause thee droplet size distribution was typially different lasers were required to fit the size distribution. The different wavelengths of light were generated using fixed wavelength laser diodes, and were ombined to pass through the shok tube on the same beam path. We used five wavelengths optimized for sensitivity to the droplet sizes expeted. The typial aerosol onentration (C v ) was between 5-20 ppmv. Figure 2.8: Shemati of Mie-sattering diagnosti for measurement of liquid volume onentration BGAS-P PHASE LASER L ABSORPTION The seond method involves measuring the gaseous fuel onentration diretly after it is evaporated. This an be done by measuring the fuel absorption of a partiular resonant wavelength (Figure 2.9). This measurement requires using two lasers of different wavelength. One wavelength is resonant and measures the fuel absorption (Figure 2.9) whih is related to the onentration through the Beer-Lambert Law. ln Equation

40 where α is the absorbane, I is the intensity of the lightt through the shok tube, I o is the inident intensity, σ is the ross setion (m 2 /mole), P is the pressure, L is the path length, R is the universal gas onstant, T is the temperature, andd X f is the ross setion. Figure 2.9: Shemati of gas phase laser absorption measurements. The seond laser is at a non-resonant absorption wavelength (Figure 2.8) and indiates the presene of liquid droplets. The non-resonant wavelength is hosen suh that theree are no gaseous absorption features in the mixture that will absorb the light. This diagnosti is then only affeted by the presene of a liquid phase. Just as in the droplet size diagnosti, the attenuation of light due to liquid droplets present in the mixture is aused by Mie sattering. This also means iff there is any ondensation on the windows this laser is affeted, whih is an indiation that the data from the other laser is suspet. So this non-resonant laser also doubles as a window leanliness monitor. Usually heating the windows resolves most ondensation problems (See Appendix A for a more detailed explanation) BFUE EL SPECTROSCOPY There is a resonant wavelength for most hydroarbons in the mid-ir, where the CH streth rovibrational band lies. This rovibrationa al band typially has three main peaks and these orrespond to the CH 2 funtional group, the CH 3 funtional group, and a ombined band. We take advantage of a transition in the HeNe laser that overlaps with this rovibrational band. There are only a few manufaturers that make very low noise 23

41 HeNe lasers with optis for transmitting 3.39 µm light. We use the Jodon TM HN-10- GIR. The laser linewidth is very narrow and near the enter of the CH streth band. For larger moleular weight moleules of interest, whih have many arbonhydrogen bonds, the multitude of rovibrational states is so dense that the spetrum is not a funtion of pressure; however, there is a temperature dependene. Raising the temperature serves to populate higher energy transitions and effetively spread the population over a wider range of energies. The result is a smoothing of the feature. More speifially the peaks derease and the tails inrease (Figure 2.11). Beause the HeNe laser sits on a point of infletion, the temperature dependene of the ross setion is relatively weak, until about 500 K where it then dereases monotonially. There is also another option for measuring fuel in the CH streth band. A differene-frequeny-generation (DFG) laser has been reently ommerialized, whih gives aess to a range of wavelengths in this band. This is partiularly useful beause the fuel spetrum an be used to measure temperature, the fuel onentration in the presene of droplets, or window fouling as in Klingbeil et al. [17] BFTIR/HEATED CELL MEASUREMENTS The measurement of ross setion must be arefully made for eah fuel that we study. We make these measurements in both heated ells and shok tubes and rossorrelate the data. In the heated ell we use a well-haraterized oven to ontrol the temperature. The fuel is evaporated into a heated tank. Beause the maximum working temperature of the pressure transduer is lower than the temperatures needed to keep the fuel in the gas phase, a N 2 buffer was oasionally required in the gas line of the pressure transduer. This avoids ondensation on the pressure sensor s head. 24

42 Figure 2.10: Experimental setup for measuring rosss setions in a heated ell with an FTIR. The ell has two sapphire windows through whih a ollimated beam from an FTIR passes. The beam is then deteted on a liquid-nitrogen-ooled MCT detetor, a Thermo Sientifi TM Model MCTA (Figure 2.10). The FTIR measures the ross setion over the entire band (Figure 2.11). These measurement ts an be made very preisely up to 300C (Figure 2.12); but in the time it takes to average over many sans (two minutes) at higher temperatures the fuel begins to deompose. One method of heking for this deomposition is to look at the integrated band strength at eah temperature. The integrated band strength is defined as the integral of the absorbane over the entire band. If this is onstant, then no CH bonds have broken and left the population. If it dereases at high temperature, then some CH bonds have been destroyed. This is furtherr evidene that the fuel annot spend a long time at elevated temperatures. Measurements an be made faster by averaging fewer sans, allowing measurements up to 500C, but result in larger error bars. To go to even higher temperatures we need to use the aerosol shok tube, itself. 25

43 Figure 2.11: Methyl deanoate spetra measured at various temperatures in a heated ell using a FTIR. Results of measurements of the CH streth band are shown in Figure 2.11 for methyl deanoate at several different temperatures. It beomes apparent that the HeNe laser at 3.39 µm wavelength does not have the largest ross setion. In experiments with low fuel loading, it would be benefiial to thee fuel sensitivity to selet a wavelength slightly above 3.4 µm to maximize the ross setion. This an be done using a DFG laser instead of a HeNe. Most of the work in this thesis uses high fuel loadings, so low absorption is not as muh of an issue, and the 3.39 µm ross setions are suffiiently strong. To summarize the data presented in Figure 2.11 for 3.39 µm, wee plot the ross setion at this wavelength versus temperature as in Figure 2.12 This plott shows the ross setions of many lowthis vapor-pressure fuels. Most fuels exhibit little temperature dependene in temperature range. This is not the ase if we extend our measurements to higher temperature using the aerosol shok tube. 26

44 Figure 2.12: Summary of ross setion measurements at 3.39 µm for various heated elll and FTIR. fuels measured using the BSHOCK TUBE MEASUREMENTS OF HIGH TEMPER ATURE CROSS SECTIONS A shok tube an aess higher temperatures using step hanges in temperature. The fuel onentration an be measured behind the inident shok,, and then the known amount of fuel an be used to measure the ross setion behind the refleted shok. First the ross-setion is measured to as high a temperaturee as possible in the heated ell / FTIR setup (Figure 2.10). Next the shok tube is filled with a fuel aerosol. A shok is initiated suh that region 2 (the onditions behind the inident shok) is at a temperature within the range of temperatures that were measured inn the heated ell, but hot enough that the fuel ompletely evaporates in region 2. Then the purely gas-phase absorbane an be measured. When the refleted shok further heats the purely gas phase mixture (region 5) the onentration is not hanged. By measuring the ratio of the absorbane of region 5 to that of region 2 (where the ross setion is known) we an alulate the ross setion in region 5 shown in the equation below (Equation 2.2). 27

45 Equation 2.22 where σ(t) is the ross setion usually measuredd in m 2 /mole at a speified temperature, and α is the absorbane (defined in Equation 2. 1) as measured in the heated ell (lower temperature) and the shok tube (high temperature). The subsripts 2 and 5 refer to region 2 (behind the inident shok) and region 5 (behind the refleted shok). This operation an be repeated for a wide range of temperatures and eah of the regions an be made to overlap to hek the self-onsisteny of these measurements, thereby measuring the ross-setionss at temperatures ranging from 300K all the way to 1200K. See Figure 2.13 and Figure 2.14 for examples of the experimental data for n- dodeane and JP-7 respetively. Figure 2.13: Cross setion measurements of n-dodeane at 3.39 µm extended to highh temperatures using the aerosol shok tube. 28

46 Figure 2.14: Cross setion measurements of JP-7 aerosol shok tube. at 3.39 µm extended to high temperatures using the These high temperature ross setions exhibit inreased satter as ompared to the low temperature measurements of ross setion. Thee unertainty of the heated ell measurements are around 3-5% %, whereas the unertainty of the high-temperature measurements is on the order of 10%. The measurements in the shok tube inorporate additional error due to the fat that at these high temperaturess the fuel begins to deompose quikly. Thus we rely on a measurement off the absorption immediately after the arrival of the refleted shok, whih is sometimes onvoluted with the shlieren spike in absorption. This added satter only affets the degree to whihh we know the high- beause we use the fuel onentration measurement from region temperature ross setions. It does not affet how welll we know the fuel onentration, 2. 29

47 2.5 27BCA ALCULATION OF EQUILIBRIUM SHOCK JUMP CONDITIONS Measurements of gas phase absorption are not all that is required to measure the fuel onentration. In Equation 2.1, the absorbane α = -ln(i/i o ) is measured in the shok experiment (Figure 2.15), σ is measured in the heated ell with the FTIR. But to find X f, we need to know T and P. Typiallyy T and P are found by measuring the shok veloity and applying the shok jump relations to find T and P in regions 2 and 5; however, to do that one needs to know the initial gas omposition or X f f. This introdues an extra unknown, and to solve this, an extra equation is required. This an be aomplished by oupling the onservation n of mass, momentum and energy equations to the Beer-Lambert equation. The solution of this set of equations iss done numerially with a program alled AEROFROSH (Appendix F) by iterating the fuel onentration to find a solution that satisfies all the equations. Figure 2.15: Example of fuel measurement behind an inident shok wave. The red trae shows droplet sattering and absorption. The blak trae shows only droplet sattering. Arrival of the inident shok ours at 0 µs and arrival of the refleted shok ours att 500 µs. Conditions: 21% O 2 in Ar with X f (n- loation is 10 m from endwall. 30 dodeane mole fration) = , T 2 = 617 K, P 2 = atm, T 5 = 993 K, P 5 = 4.23 atm. Measurement

48 BVALIDATION OF AE EROFROSH AEROFROSH was tested against analytial solutions made by Ghua et al. [18] and Marble et al. [19] for a water aerosol being evaporated by a shok wave. Figure 2.16 shows the temperature behind an inident shok whose strength is suh that the ratio of pressure behind the shok to the initial pressure is 3. This plot was made to repliate Figure 4 in Ghua s work [18]. The initial quality of the water wass varied to understand the effet on the post-shok temperature. Up to 6% (byy mass) liquid, there is an inverse relationship between post shok temperature and quality. This is due to the enthalpy lost to the heat of evaporation. However, at higher qualitiess the post-shok temperature stays onstant with inreasing quality, beause here the post-shok gas phase is in equilibrium with the liquid phase at the saturated vapor pressure, and as a result droplets remain. AEROFROSH alulations agree with Ghua within thee operationall range of the aerosol shok tube. Also at the limit of omplete evaporationn AEROFROSH agrees with the alulations in Ghua et al. [18] and Marble et al. [19]. Figure 2.16: Comparison of AEROFROSH alulations to those made by Guhaa in 1992 [18] and Marble made in 1968 [19] for a shok traveling in pure water of varying initial quality. 31

49 AEROFROSH was also used to predit the onditions in the aerosol shok tube. To validate the alulations, we measured pressure near the endwall, and ompared the measured pressure to the alulated pressure. The measurements mathed the alulations to within the measurement error of 3% (Figure 2.17). Figure 2.17: Comparison of measured pressure (2 m from the endwall) with AEROFROSH those predited by To inrease the sensitivity of our validation target,, temperature was measured using two methods. The first was done using two DFG lasers that probed two different wavelengths in the CH streth band for n-dodeane. The pair hosen had a ratio of ross-setionss that was dependent on temperature. This enabled a sensitive measurement with an error in temperature of 5% [17]. Figure 2.18 shows the agreement using this measurement tehnique is within the measurement error. It is, however, slightly disonerting that at high temperatures we see that the error is not random and has a systemati offset. We believee that this is due to slight deomposition of the fuel. This prompted further study using a more stable moleule for this temperature measurement t. 32

50 Figure 2.18: Comparison of measured temperature using a two-wavelength fuel diagnosti with alulated temperature. The seond method of temperature measuremen nt involved using a two-line CO 2 diagnosti, whih is desribed in Ren et al. [20]. This method wass muh more aurate with an error of only <2% in temperature. Within the auray of the measurement, the alulations of onditions are orret using AEROFROS SH (Figure 2.19). Figure 2.19: Comparison of measured temperature using a two-line CO 2 tehnique with alulated temperature. 33

51 Region 2 Region 5 AEROFROSH Measured Error AEROFROSH Measured Error K K K K K K Table 2.1: Tabulated results from measurement of temperature using the CO 2 diagnosti. Results show good agreement between measured temperature and predited temperatures using AEROFROSH BAPPLICATIONS OF MEASUREMENTS BEHIND REFLECTED SHOCK WAVES One we are onfident that the alulated onditions behind the inident shok are orret, we an easily apply the shok-jump equations to solve for the onditions behind the refleted shok wave (assuming stagnated flow in this region). Then, with welldefined onditions, high quality ombustion hemistry measurements behind the refleted shok an be made. The main appliation of this faility is in providing validation targets for detailed hemial kineti models. These an take the form of ignition delay times, emission, and/or speies-time history measurements. Ignition delay times provide a single number to test simulations of the entire mehanism. The ignition delay time is defined as the time in whih it takes a ertain gas phase mixture to auto-ignite at a given temperature and pressure. Beause this time depends on all the intermediate reations it is a good metri for mehanism validation. The ignition event is haraterized by an exponential rise in the reation rate governed by a thermal or radial hain reation. This auses an exponential rise in the produts, OH and CH as well as a rapid onsumption of hydroarbons. For energeti mixtures the pressure and temperature will rise very sharply. 34

52 Figure 2.20a,b: a. Plots of pressure versus time measured 2 m from the endwalll in AST I for mixtures of n-dodeanee / 21% O 2 / Ar at various temperatures = 1.0, P = 6.0 atm. b. Summary of ignition delay times plotted on an Arrhenius plot. Shown in Figure 2.20a are the pressure traes for a mixture of n-dodeane and 21% O 2 in Ar at various temperature. Eah of these experiments s an be plotted on an Arrhenius plot to show more learly the temperature dependene (Figure 2.20b). In this ase we used pressure to indiate ignition, but OH* and CH* emission an also be used. A omparison of the measured ignition delay time to the mehanism will reveal the suess or failure of a kineti model. The relative behavior of the model with hanges in initial temperature and pressure an give insight into whih elementary reations or reation pathways are missing or in need of adjustment. The ignition delay time is very useful; however, it is only one parameter, and with a model that has so many parameters it is not enough to onstrain the mehanism ompletely. Therefore we an also measuree speies time histories and ompare those to the simulated speies time-histories produed using the mehanism. These measurements provide an aount of the evolution inn time of the onentration of a ertain moleule in the gas mixture. The omparisons are most useful when omparing a moleulee that is involved in many of the key reations of a mehanism. In some ases, these measurements provide the basis for a fitting sheme where the most sensitive reation rates are perturbed to enfore fitting between thee model and the measurements. 35

53 2.7 29BCONCLUSIONS In this hapter we looked at the details of how the aerosol shok tube works, and how it an be used to make very useful measurements of the ombustion hemistry of low-vapor-pressure fuels. The aerosol is generated using ultrasoni nebulizers and has a size distribution typially between 1-10 µm. The aerosol an be entrained in an oxidizing bath gas and flowed into the shok tube. The inident shok wave is used to evaporate and mix the fuel, and at this stage laser absorption is used to probe the CH streth rovibrational band to measure the purely gas phase absorbane. With that information and the shok speed we use AEROFROSH, whih ouples the shok jump equations to the Beer-Lambert relation to alulate the onditions and fuel mole fration. We verified that this alulation proedure mathes measurements, ensuring the quality of these measurements. This method is shown to be a very useful tool for the study of hemial kinetis. However, the aerosol uniformity, whih has not been disussed in great detail, will be disussed in the next hapter. 36

54 Chapter 3: 7BUNIFORMITY IN THE AEROSOL SHOCK TUBE The improved, seond-generation aerosol shok tube (AST II) has been developed for the study of the hemial kinetis of low-vapor-pressure fuels. These improvements enable a wider range of fuel onentrations and enhaned spatial uniformity relative to our initial aerosol shok tube design (AST I). In addition, the design of AST II limits the aerosol loading zone in the shok tube to a fixed region (1.2 m in length adjaent to the shok tube endwall). AST II ahieves these improvements by using a separate holding tank to prepare the aerosol mixture and a slightly under-pressure dump tank to arefully pull the aerosol mixture into the tube in a plug-flow. This filling method is apable of produing room temperature test gas mixtures of n-dodeane with equivalene ratios of up to 3.0 in 21% O 2, three times the loading ahievable in the earlier AST I that used a flow-through strategy. Improvements in aerosol uniformity were quantified by measuring the liquid volume onentration at multiple loations in the shok tube. The measurements made over a length of 1.1 m of shok tube indiate that the AST II method of filling produes non-uniformities in liquid volume onentration of less than 2%, whereas in the AST I method of filling the non-uniformities reahed 16%. The 37

55 improved uniformity an also be seen in measurement of gas phase fuel onentration behind the inident shok wave after the liquid droplets have evaporated. Signifiant redution in the satter of ignition delay times measured using AST II have also been ahieved, onfirming the importane of uniform loading of the aerosol in making highquality ombustion measurements BINTRODUCTION Although shok tubes are normally used to study purely gas-phase phenomena, they have also been used to study aerosols [21-31]. The primary hallenge, however, in studying aerosols in shok tubes is in ahieving a spatially uniform distribution of aerosol test mixture. A spatially non-uniform aerosol mixture an degrade the quality of the shok tube data; in earlier studies [21-31], non-uniformity in the spatial distribution of the aerosol tended to ompliate the interpretation of the measurements and the determination of inident and refleted shok test onditions. The uniformity of the final (post-evaporation) gas mixture is highly dependent on the spatial uniformity of the initial aerosol onentration. High levels of uniformity are easily ahieved in gaseous mixtures beause gases oupy the entire volume. However, an aerosol/gas mixture does not behave in this ideal manner. During filling, the gas tends to aelerate faster than liquid droplets, leaving the droplets behind and thus reating a spatially non-uniform aerosol loading. Smaller droplets will trak the gas flow better, but small diameters render high fuel loadings diffiult. A balane must be met between the benefit of high fuel loadings as gained from using larger droplets and the benefit gained by having smaller droplets whih trak the gas-phase flow and more easily form uniform spatial distributions. Previous work has found this balane in droplets reated by ultrasoni nebulizers with mass mean diameters of around 4-5 µm [21, 22]. Other researhers have also been onerned with filling a tube with aerosol. Some have used a diret-injetion aerosol filling strategy. One suh method is to use a point soure or soures where fuel injetors spray the fuel diretly into the shok tube either prior to arrival of the inident shok, after the inident shok or after the refleted 38

56 shok [24-28]. However, these methods produe extremely non-uniform onditions due to the fat that the spray omes out in a one pattern and there are regions in the tube that are not filled ompletely. Another method is to prepare a mixture of aerosol in a separate volume, after whih the aerosol mixture an be expanded into the shok tube through a tube onneting the two volumes with a valve to ontrol the filling proess [29-31]. These methods are better suited for smaller partile sizes, on the order of sub-mirons, beause the high aeleration of the gas at the small area orifie will result in signifiant lagging of larger diameter partiles. No quantifiation of the spatial uniformity was found in the studies disussed above; however in ref. [31] a uniformity of around 20% rms was extrated from time-resolved drum-amera measurements. There also have been a number of studies in whih a uniform mixture of aerosol is reated in a ontinuous flow sheme. In these shemes the aerosol is generated in a moving stream of gas and arried with the gas flow. The gas flow is made to pass turbulene-generating strutures in order to mix the aerosol; this approah is ommonly used in aerosol wind tunnels. In one study, ref. [32], non-uniformities as low as 11% over a 30x30 m ross-setion of the wind tunnel for aerodynami droplet sizes of 10 µm were reported. Using a similar tehnique, Brown was able to ahieve 10% variation in aerosol onentration [33]. At very low pressures of 1-10 torr, other researhers have tried with partial suess to minimize spatial non-uniformities [34, 35]. These ontinuous-flow filling tehniques have also been used in shok tubes, For example, the first-generation aerosol shok tube, built in our laboratory, used a ontinuous-flow filling tehnique [21, 22, 36]. In this faility, the aerosol is reated with an ultrasoni disk nebulizer submersed in a liquid fuel pool. A ontinuous flow of gas is established over the nebulizer, into a manifold, through the shok tube endwall, through the entire length of the shok tube driven setion, and into a mehanial pump. The aerosol arried by this flow of gas into the manifold is aelerated through a series of slightly opened poppet valves in the shok tube endwall. The narrow passages of the poppet valves reate turbulene whih mixes the aerosol, reating a uniform aerosol near the endwall. As the mixture flows further away from the endwall the turbulene 39

57 dissipates and the aerosol tends to settle to the bottom of the tube resulting in nonuniformities. The uniformity has been measured using a light sheet and imaging the sattering at 90 degrees. The best uniformities seen with this method (variations of 24%) were found at the greatest observation distane from the shok tube end wall (45 m) in the study. Results from gas phase absorption of the fuel behind the inident shok indiate that this non-uniformity is lessened by the proess of evaporation and diffusion, beause variations as small as only 1.7% were reported using the same filling tehniques [21]. The goal of the urrent study is to develop a new method to introdue aerosol into the shok tube, with higher spatial uniformity (partiularly over the last meter of shok tube length nearest the endwall), and with the flexibility to fill the shok tube with a wider range of aerosol loadings BAEROSOL SHOCK TUBE GENERATION II BEXPERIMENTAL APPARATUS A simplified shemati of the experimental setup is shown below in Figure 1a. The shemati shows a ross-setion of the shok tube as viewed from the side. The driver setion of the shok tube is separated from the driven setion (8.4 m long and 10 m square ross-setion) by a thin polyarbonate diaphragm. The test setion (1.2 m long and 10 m square ross-setion), whih ontains the aerosol, is separated from the rest of the driven setion by a gate valve. The gate valve was speially designed to make a smooth interior when opened, and when losed restrits the aerosol from filling the entire driven setion of the shok tube. Another gate valve at the end of the test setion ats as the endwall when losed. When open, the endwall gate valve allows aerosol to flow into the test setion from the aerosol mixing plenum (50 m long and 24 m diameter). The 83 liter dump tank is onneted through a 5.5 mm orifie and ball valve to the test setion and provides expansion volume as the aerosol is pulled into the test setion. 40

58 The proedure for filling the shok tube in four steps is shown in Figs. 1a-1d. First, the gas pressures in all volumes are set (with all valves losed) with a mixture of oxidizer and diluent. The aerosol mixing plenum and the shok tube test setion are set to pressure P A. The pressure P B in the dump tank is set lower than P A. The pressure in the rest of the driven setion is set to the final desired pressure P 1 suh that when the ball valve to the dump tank is opened and the pressures in the other volumes equilibrate, all pressures are equal to P 1. (Figure 3.1a) The equations for alulating the initial pressures are shown in Equations 3.1 and 3.2 Seond, the aerosol mixture is reated. Inside the aerosol mixing tank a fan is turned on that mixes the entire volume. The fan rotates at approximately 100 rpm and has six individual blades equally spaed along the axis of rotation with an overall blade diameter of 12.5 m. An ultrasoni nebulizer array is then turned on generating liquid droplets whih are quikly aught up in the mixed flow. The fan and nebulizers are left on until the liquid loading in the mixed gas reahes the desired level. Then the nebulizers are turned off. (Figure 3.1b) Some settling of large drops may then our. Third, the aerosol mixture is transferred into the test setion of the shok tube. First, the endwall gate valve is opened, and beause the pressure differene aross this valve is zero, there will be no flow. Immediately following this, the ball valve is opened and gas begins to flow from the test setion into the dump tank. As gas from the test setion flows from the test setion into the dump tank, the mixture of aerosol is pulled in as a plug flow into the test setion. The pressures (P A, P B, and P 1 ) are set suh that the dividing surfae between the aerosol mixture and the non-aerosol gas propagates well past the ball valve into the dump tank. (Figure 3.1) Finally, the pressures in all the volumes equilibrate, to a value suh that there is no pressure differene aross the driven setion gate valve. This valve an then be opened without any resulting flow; the endwall gate valve and ball valve are then losed. The result is a smooth-walled shok tube with a spatially uniform aerosol near the endwall. (Figure 3.1d) The shok wave experiment an then be initiated, typially within a few seonds. 41

59 Figure 3.1a-d: (a) Diagram illustrating omponents of AST II and initial pressures set for operation. (b) Aerosol mixture being prepared in the mixing plenum. () Endwall gatee valve and ball valve opened and aerosol mixture expanded into the test setion. (d) Endwall gate valve and ball valve losed and the driven gate valve opened; the shok tube is ready for initiation of the shok wave. 42

60 BCAL LCULATION OF LOADING PRESSURES The initial pressures in the shok tube and tanks must be set suh that one equilibrated, the desired final pressure in the shok tubee is reahed.. This final pressure, P 1, is the pressure of the gas thatt the inident shok wave will propagate into and is used in alulating the onditions behind the refletedd shok wave. A shemati demonstrating the filling proesss and the relevant volumes is shown in Figure Figure 3.2: Shemati of filling proess for seond generation aerosol shok tube. Initial pressures in the test setion (V TS ) and mixing tank (VMT) are both PA and are separated from the dump tank (V B ) by a ball valve, whih is at a lower pressure P B. When the ball valve is opened the pressures equalize to P 1. At this point the expansion is suffiient for the gas from the mixing tank to fill the test setion. The olors represent the boundaries of the respetive volumes. The dump tank volume is denoted V B and has a low initial pressuree P B. The test setion volume is labeled V TS and is at a higher initial pressuree P A. Thesee volumes are separated by a ball valve. The aerosol mixing tank volume is V MT and also has an initial pressure P A. The volume V is the displaed volume, whih is the sum of V TS and the partial volume of the dump tank that is filled with the test gas. This displaed volume is an independent parameter, whih an be varied depending on how far into the shok tube the aerosol needs to fill. We an then define X as the non-dimensional expansion parameter. X = V /V TS Equation

61 To fill only the entire test setion, X would equal one. In pratie, we employ a larger fill parameter X=1.5 to ensure omplete filling. Our goal is to speify the initial pressures P A and P B based on a desired P 1 and a fixed X=1.5. The expansion proess an be assumed to our isothermally, whih allows us to write Boyle s Law for the test gas. P A V MT = P 1 (V MT + V ) Equation 3.2a By dividing Equation 3.2a by V MT we an solve for P A. We an also substitute Equation 3.1 for V to put Equation 3.2a in terms of X. P A = P 1 (V MT + V TS X)/V MT Equation 3.2b Now, given a desired final pressure P 1 and expansion parameter X, Equation 3.2b speifies the initial pressure in the test setion and the mixing tank (P A ). To find the initial pressure in the dump tank (P B ) we an write Boyle s Law for the entire system. P A (V TS + V MT ) + P B V B = P 1 (V TS + V MT + V B ) Equation 3.3a Sine we know P A from Equation 3.2b, we an rearrange Equation 3.3a to solve for P B in terms of known quantities. P B = (P 1 (V TS + V MT + V B ) P A (V TS + V MT ))/V B Equation 3.3b Using this model, we an generate urves for P A and P B based on our desired P 1. In Figure 3.3, X = 1.5 and the volumes used were V TS = m 3, V MT = m 3, and V B = m 3, 44

62 Figure 3.3: Calulation of initial pressures (P A and P B ) based on desired final pressure (P 1 ) using an isothermal assumption. Equation 3.2b and Equation 3.3 do a reasonablee job of prediting the expansion behavior during the fill proedure of aerosol shok tube; however, in pratie a more aurate model is used to set the initial pressures based on a hybrid isothermal expansion and isentropi ompression model, whih is detailed in Appendix D. Based on this simple analysis it beomes apparent that to avoid high h mixing tank pressures (P A ) it is neessary to make the dump tank large. The volumes used in this work allow for P 1 s up to ~442 torr while limiting P A to atmospheri pressuree (a limit that is imposed by the dump tank design). If higher pressure experiments are desired, one may use dump tank that an withstand higher pressures or inrease the dumpp tank volume (V B ). 3.3 RESUL 32BR TS Three aspetss of the AST II aerosol loading sheme performane are disussed: first, the behavior of the nebulizers; seond, the spatial uniformity ahieved in the aerosol mixing plenum; and third, the spatial uniformity in the filled shok tube volume BNEB BULIZER PERFORMANCE The effetiveness of the AST II method of loading the shok tube with aerosol in a spatially uniform manner over a wide range of volumetri onentrations depends 45

63 strongly on the size of the liquid droplets that are used. A high liquid-volume fration an be ahieved by inreasing the number of droplets or the size of droplets. Beause volume is strongly dependent on diameter, it is advantageous to use larger diameter droplets. However, larger droplets settle faster and lag in gas flows. So a balane must be struk while onsidering the limitations of the aerosol generators. The ultrasoni erami-disk nebulizers are ideal for this purpose beause they work even with visous fuels suh as diesel (DF-2) and bio-diesel surrogates suh as methyl deanoate, and they are apable of produing high liquid volume frations with suffiiently small diameters that trak the gas flows that our with this filling method. In order to ensure onsistent results the droplet size distributions of nebulized aerosols were measured for various liquids. The nebulizer was operated in a tube with an inner diameter of 12 m (in liquid depths above the erami-disk of m) and the droplets were entrained in a flow that went from the sidewall of the tube and out of the top (see Figure 2.3). Aerosol size distributions were measured using a Malvern Sprayte TM (Model RTS5214). Figure 2.4 shows a plot of the droplet size distribution for several liquids: water, n-dodeane, and diesel (DF-2) fuel. The flow rates were low to keep with the low veloities enountered when filling of the shok tube. The size distributions were found to be nearly the same for all three liquids. The mass-averaged size was between 4-5 µm. The Stokes settling time for this droplet size is approximately 1 minute BMIXING TANK LOADING UNIFORMITY Spatial uniformity and knowledge of loading levels in the mixing tank is of ruial importane. This an be quantified using laser extintion if we take advantage of our measured droplet size distribution. The onentration of the aerosol an be related to the attenuation of a monohromati laser beam using Mie theory. The theory provides that the extintion ( -ln(i/i o ) ) is given by this equation [37] 46

64 -ln (I/I 0 ) = C V { 0 Q ext (d, n( )) f(d) /4 D 2 L dd }/{ 0 f(d) /4 D 3 dd } = [m -1 ] L [m] CV [ppmv] Equation 3.4 where the I is the intensity of the light that is transmitted, I o is the inident intensity, L is the path length, C v is the liquid volume onentration, Q ext is the Mie oeffiient whih depends on the droplet diameter (D) and the index of refration whih is dependent on the wavelength of light (λ), and f(d) is the frational size distribution. Thesee integrals an be omputed for the measured f(d) shown in Chapter 2 (Figure 2.4) and theoretial values for Q ext. Beause the size distributions are similar and the indies of refration are nearly the same (n= ) for the liquids used in this study, the integrals an be simplified as shown in Equation 3.4. The laser setup (shown in Figure 3.4) on the mixing plenum onsisted of two lasers 30 m apart whih were sent aross the enter of the plenum in a horizontal plane (L=24.2 m). The windows were heated (~40 C) to avoid ondensation. Figure 3.5 below shows a typial test result. Figure 3.4: The optial setup on the mixing plenum to measure thee uniformity and the loading. 47

65 Figure 3.5: Uniformity in tank: water aerosol. Two lasers at 10 and 40 m from the endwall. The disrepany between the two measurementss indiate the magnitude of the non-uniformities present in the mixture at a partiular time. The oeffiient of variation (COV=StDev/Ave) is plotted as a funtion of time. Large variations are seen in the onentration signall while the nebulizer is on as the jets above the nebulizer disk steer the laser beams away from the detetors. The onentrationn rises quikly while the nebulizerr is on, and then one the nebulizer is turned off (at 4s) these variations are redued. When the fan is turned off (at 6 s) the variations are redued further, and we see a veryy slow drop in the onentration (around 1%/s). This is due to the larger droplets settling. This effet is less evident in atual experiments beause the fill happens diretly after the fan turns off. The oeffiient of variation (COV) for these measurements showss that the non-uniformity in the tank is around 1%. No strong dependene on fan speed (near the nominal value of 100 rpm) was seen. 48

66 Figure 3.6: Filling the tank to different onentrations: water aerosol. Various urves represent different lengths of time that the nebulizer and mixing fan was turned on.. This shows the effets of leaving the nebulizer running for different amounts of time and the range of ahievable onentrations (1-100 ppmv). Figure 3.6 shows the effets of leaving the nebulizer running for different amounts of time and the range of ahievable onentrations (1-100 ppmv). The AST II onfiguration signifiantly extends the onentration range beyond that whih was previously ahieved with the AST I method of filling thee shok tube (5-20 ppmv) BSHO OCK TUBE LOADING UNIFORMITY Studies of shok tube loading uniformity were performed using a Plexiglass mok tube to aid in visualization. Loading in the mok tube was studied using three lasers stationed at different distanes away from the endwall (10 m, 60 m, 110 m) rossing the tube at its mid setion. The windows were heatedd (~40 C) to avoid ondensation. See Figure 3.7 for the experimental setup and Figure 3.88 for the results of a typial filling experiment using the AST II method. 49

67 Figure 3.7: Optial setup for measuring non-uniformities in AST II method of filling. Figure 3.8: Plot of laser extintion measurements at three loations along the tube and the resulting oeffiient of variation: water aerosol. AST II generatess non-uniformities muh smaller than AST I, in this ase 2%. 50

68 This filling experiment (Figure 3.8) was onduted with a water aerosol using loading pressures of P A =759 torr, P B =326 torr, P f =460 torr, and X=1.5. The mixing plenum filling time sequene was as follows: gas flow and fan in the mixing plenum turned on at -10 s, nebulizer is turned on at -8 s and turned off at -5 s, fan turned off at -3 s. This filling sequene ahieved an aerosol liquid loading of 44 ppmv (see solid blak urve). The mok tube filling sequene was as follows (same time sale): ball valve to the dump tank and gate valve to mixing plenum opened at 0 s, the aerosol ontat surfae rosses the first laser almost immediately, the seond laser registers an inrease in the aerosol onentration at 1 s, and the third laser registers an inrease at 2.5 s. The liquid volume onentrations derease between 0 s and 4 s due to the expansion of the mixture; one the pressures equalize at 4 s the onentrations at all three laser measurement loations are very similar in magnitude and derease at a slow rate due to settling. The COV time-history indiates that the non-uniformities remain less than 2% after filling. An idential set of measurements were made using the AST I flow-through method of filling and non-uniformities were signifiantly higher in all ases and typially around 16%. In order to minimize the non-uniformity using the AST II, one very important parameter, the plug flow rate, must also be optimized. The next set of plots in Figure 3.9- and Figure 3.10 show the effet of varying the fill rate. 51

69 a b Figure 3.9a-: These plots show measurements of aerosoll onentrationn at three loations while filling the shok tube using the AST II filling method (a) 50 slpm fill rate (b) 4.0 slpm fill rate () 1.0 slpm fill rate. The flow rate was varied by using various diameter orifies. In Figure 3.9a the flow rate is high (50 slpm) and the aerosol takes only 0.2 s to fill the test setion; large differenes in the aerosol onentration are seen at the three laser measurement loations. This variation is likely due to the separation of the liquid droplets from the bulk gas at high flow veloities. In Figure 3.9b the test setion is filled in 2.7 s (4.0 slpm) and the aerosol is muh moree uniform. At signifiantly slower flow 52

70 rates (1.0 slpm), as in Figure 3.9, the aerosol again beomes less uniform; droplets near the aerosol ontat surfae evaporatively ool and settle, beginning a river-like flow. At these filling rates, the aerosol flow is not an ideal plug flow and oupies only the bottom half of the tube s ross-setion. The region of uniformity ours when the flow rate is between 2 and 5 slpm or at veloities between m/s. The optimal flow rate is obtained by venting the shok tube test setion (or mokk tube in the visualization studies) through a throat or orifie just before the ball valve to the dump tank. The throat that was hosen to produe a veloity in the test setion of 45 m/s has an orifie diameter of 5.5 mm. Figure 3.10 shows many filling experiments where the flow rate was varied by opening the ball valve to different angles. Figure 3.10: The relationship between the resulting non-uniformity to the flow rate with whih the tube was filled. The flow rate was varied by using various diameter orifies. The optimal flow rate range is between 2 and 5 slpm (gas veloity ~45m/s). 3.4 SHOCK 3BS WAVE EXPERIMENTS BFUEL LOADING UNIFORMITY After the shok tube filling is omplete, a shokk wave is initiated. The inident shok wave propagates through the driven setion of thee shok tube into the aerosol filled test setion. For these experiments a fuel aerosol (e.g. dodeane, diesel fuel, et.) is used in order to study the hemial kinetis of that partiular fuel. When the shok passes through the aerosol it quikly aelerates and evaporates the fuel. The evaporated fuel mixture then flows toward the endwall behind the inident shok. 53

71 Post-evaporation gas-phase uniformity in the test gas mixture using the new aerosol loading strategy was measured using laser absorption in region 2. In region 2 the fluid is flowing toward the endwall and a time-resolved laser measurement at a single loation gives information about all of the fluid elements that pass the laser. Figure 3.11 below shows the optial setup used in these measurements. Figure 3.11: Top View: Diagnostis used in aerosol tube experiments. Pressure measurements are used to measure shok speed and ignition delay time. Laser light at 650 nm an be used to measure liquid aerosol onentration. Laser light at 3.39 µm is used to measure gas phasee onentrationn after the fuel has evaporated. Side View: Inident shok propagates into aerosol evaporating and mixing fuel and oxidizer. This region is alled region 2. 54

72 Shown in Figure 3.12a-, are extintion measurements ( ln(i/i o ) where I o is the inident laser light intensity and I is the intensity after being attenuated through the shok tube) for experiments in the shok tube. At this measurement loation (10 m) two laser beams with different wavelengths were pithed through the shok tube. The visible, 650 nm laser beam was attenuated solely by Mie sattering when there are liquid droplets present. The mid-infrared 3.39 µm laser beam was also attenuated by Mie sattering, but in addition is resonant with the C-H streth vibrational mode found in gaseous hydroarbons in the mixture. The visible beam provides information on the aerosol loading and onfirms when the aerosol is fully evaporated (i.e. no attenuation of 650 nm at ~100 µs). The mid-infrared beam provides purely gas-phase fuel onentration after the aerosol is fully evaporated. This is done using a method desribed in Davidson et al. [21]. 55

73 a b Figure 3.12a-: Three shok experiments highlighting region 2 post-evaporation uniformity for a range of fuel loadings alll with n-dodeane in 21%O2/Ar. (a) Conditions: X f (n-dodeane mole fration) = , T2 = 617 K, P 2 = 1.11 atm, T5 = 993 K, P 5 = 4.23 atm (b)) Conditions: X f = , T 2 = 584 K, P 2 = 1.64 atm, T 5 = 921 K, P 5 = 6.25 atm () Conditions: X f = , T 2 = 534 K, P 2 = 1.62 atm, T 5 = 838 K, P 5 = atm. 56

74 Figure 3.12a- shows these laser extintion measurements in regions 1, 2, and 5. Before the arrival of the shok wave, in region 1, there is no hange in the extintion (- ln(i/i o )) of both wavelengths beause the aerosol is stagnant. After the inident shok arrives, the mixture is ompressed and the extintion rises. In the 650 nm measurement the signal quikly reahes a peak and falls to zero due to the rapid evaporation behind the inident shok. The 3.39 µm measurement falls due to evaporation as well, but then asymptotes to a steady plateau value. This plateau value of the gas phase absorption is used to alulate the fuel onentration. This measurement also indiates that the test gas mixture passing the measurement loation had a very uniform fuel loading for all three tests shown here (<1% variation). This is a signifiant improvement over the results obtained using the AST I filling method where the non-uniformities were typially >5% and never better than 1.7% [21]. At the arrival of the refleted shok the mixture is further ompressed. As a result the attenuation inreases as seen in Figure 3.12a. A shlieren spike is also seen in the laser measurements, where the beam is temporarily steered off the detetor beause of the passage of the refleted shok. In Figure 3.12b the inreased attenuation is not seen after the refleted shok (owing to a derease in fuel onentration that ours during the shlieren spike), and in Figure 3.12 the attenuation falls prior to the arrival of the refleted shok. In both ases (Figure 3.12b and Figure 3.12), the arrival of the aerosol ontat surfae prior to the arrival of the refleted shok at the measurement loation aused a derease in the absorption. In order to avoid the influene of the aerosol ontat surfae on the measurement, either the measurement loation should be moved loser to the endwall or the test setion (whih is initially filled with aerosol) should be made longer. Cases like Figure 3.12b and Figure 3.12 are important in haraterizing the faility to understand the effets of the aerosol ontat surfae. Then when making ombustion measurements behind the refleted shok wave, the measurement loation an be seleted nearer the endwall to avoid any non-uniformities in onentration assoiated with the aerosol ontat surfae. 57

75 BIGNITION DELAY TIME MEASUREMENTS As an example appliation and demonstration of this new faility to study ombustion hemistry phenomena, ignition delayy times in n-dodeane/o 2 /argon mixtures were measured. Ignition delay time measurements are used to desribe the global ombustion performane of fuel/oxidizer mixtures. The ignition delay time is the length of time that a mixture, at a given initial pressure and temperature (typially under onstant volume and energy onstraints) will take to ignite; ; measurements an be made diretly from real-time pressure profiles. These AST II measurements are ompared with previous measurements using AST I [21] in Figure Figure 3.13: Ignition delay times for AST I [21] and ASTT II. AST II data produes signifiantly redued satter and slightly lower mean values than AST I data. Pressure and Φ (equivalene ratio) were normalized to 5.0 atm and Φ=0.5 using P and Φ dependenes, respetively. The new aerosol filling method produess redued satter. The AST II points are slightly shifted to shorter ignition times. A omparison analysis of the AST I data shows evidene of larger non-uniformity of fuel loading, suggestingg that a signifiant portion of 58

76 the measurement differene was due to non-uniformities in the initial aerosol mixture. The new filling method produes a more uniform initial aerosol mixture and signifiantly redues the satter in the ombustion measurements that were made BCONCLUSIONS A seond-generation aerosol shok tube has been developed. Three improvements over the previous design were demonstrated. The new method now onstrains the amount of shok tube that is filled with aerosol, produes a muh more uniform aerosol, and allows filling the shok tube with muh higher onentrations of fuel. These improvements will enable more aurate measurements of ombustion hemistry proesses, partiularly as are needed for pratial low-vapor-pressure fuels. 59

77 Chapter 4: 8BAPPLICATION 1 - IGNITION DELAY TIMES Gas-phase ignition delay times were measured behind refleted shok waves for a wide variety of low-vapor-pressure fuels. These purely gas-phase measurements of ignition delay times, without the added onvolution with evaporation times, were made possible by using the aerosol shok tube methodology. The fuels studied inlude three large normal alkanes, n-deane, n-dodeane and n-hexadeane; one large methyl ester, methyl deanoate; and several diesel fuels, DF-2, with a range of etane indies from 42 to 55. The refleted shok onditions of the experiments overed temperatures from 838 to 1381 K, pressures from 1.71 to 8.63 atm, oxygen onentrations from 1 to 21%, and equivalene ratios from 0.05 to Ignition delay times were measured using sidewall pressure, IR laser absorption by fuel at 3.39 µm, and CH* (at 431 nm) and OH* (at 306 nm) emission. Measurements are ompared to previous studies using heated shok tubes, where available, and urrent models. Model simulations show similar trends to the urrent measurements exept in the ase of n-dodeane/21% O 2 /argon experiments. At higher temperatures, e.g K, the measured ignition delay times for these mixtures 60

78 are signifiantly longer in lean mixtures than in rih mixtures; urrent models predit the opposite trend. As well, the urrent measurements show signifiantly shorter ignition delay times for rih mixtures than the model preditions BINTRODUCTION The aerosol shok tube has been shown to provide a unique method of experimentation for low-vapor-pressure fuels that effetively separates the evaporation proess from the ombustion. The AST II method of filling provides a homogeneous initial aerosol distribution in the shok tube. This along with a well-validated method of determining the onditions behind the inident and refleted shok waves make it ideal for measuring ignition delay times for low-vapor-pressure fuels. Ignition delay times are good indiators of the overall behavior of ombustion reations and are regularly used as performane benhmarks for detailed hemial mehanisms. Validation of the detailed hemial mehanisms is very important, beause these mehanisms are obliged to use many reations and rate onstants that are only estimated or theoretially predited and are not derived from experiment. For large fuel moleules (whih typially have low vapor pressures), the number of speies that are formed during deomposition and oxidation an be very large. For example, a Lawrene Livermore National Labs (LLNL) hemial mehanism for alkanes up to n-hexadeane (C 16 H 34 ) ontains over 2,000 speies and over 8,000 reations [43]. Ignition delay times provide useful targets and some onstraint on the modeling of these omplex systems. Muh work is still needed to fully understand the ombustion of these large fuel moleules beause of the inreased omplexity of these mehanisms, and the added diffiulty in performing experiments with these fuels [54]. This is unfortunate, beause most of the devies that utilize ombustion are powered by low-vapor-pressure fuels inluding diesel and bio-diesel fuels, and roket and jet fuels. Many new mehanisms have been developed reently [43-46, 61, 64], but more experimental validation targets are required for these fuels and their surrogates [62]. To help address this need, we have studied a variety of low-vapor-pressure fuels using the aerosol shok tube methodology, 61

79 providing, in many ases, the first purely gas-phase shok tube ignition delay time measurements for these fuels BLOW-VAPOR-PRESSURE FUELS Fuels used in ombustion devies are rarely omposed of a pure single omponent, and in order to model a real, multi-omponent fuel, single- or multiomponent surrogate mixtures are often employed [47, 62]. This motivates the study of various types of pure fuels that display harateristis similar to real fuels, in addition to the study of the real fuels themselves. Beause fuels suh as jet fuel, diesel, and biodiesel are mostly made up of large moleules, the surrogate omponents whih best represent these fuels have low-vapor-pressures; the aerosol shok tube is thus well-suited for the study of these fuels. Here we fous on n-alkanes, whih an make up a signifiant fration of distillate fuels, diesel fuels themselves, and methyl deanoate, a large methyl ester that is struturally related to bio-diesel BN-DECANE There are many ignition delay time studies of n-deane [48-51, 60, 12]. This is in part beause its vapor pressure at room temperature is near 1.4 torr and this is suffiient to make a purely gas-phase mixture at relevant mixture frations. n-deane (C 10 H 22 ) has been ommonly used as the alkane representation for jet fuel-relevant surrogates, but has also been used for diesel [59] and bio-diesel [45] omparisons BN-DODECANE Less abundant are ignition delay time studies foused on n-dodeane [48, 52]. The vapor-pressure at room temperature of n-dodeane is lower at 0.13 torr. This limits the fuel loadings that an be ahieved, and heating is usually neessary to perform shok tube experiments. n-dodeane (C 12 H 26 ) has also been used as the alkane omponent in jet fuel surrogates. 62

80 BN-HEXADECANE There are very few studies on the purely gas-phase oxidation of n-hexadeane [13, 53, 14] beause its vapor pressure at room temperature is only 1.4 mtorr. n- hexadeane (etane) is a major alkane onstituent of diesel fuel, and appears to be very important in influening its behavior upon oxidation. n-hexadeane (C 16 H 34 ) is one of the speies that is used in the definition of the etane index sale to haraterize the ombustion harateristis of diesel fuel. By definition n-hexadeane has a etane index of BDIESEL (DF-2) Diesel fuel is a low-volatility distillate fuel most widely used in ombustion devies. Heated shok tubes have been used in the past to study diesel ignition [55, 56], but with a distillation urve that extends up to 350C, it an be a very diffiult task to get all of the heavier omponents of the fuel mixture into the shok tube. Here we examine several different diesel fuel samples to show how variability in omposition and etane index affet ignition delay times BMETHYL DECANOATE The author is unaware of any purely gas-phase shok tube studies of methyl deanoate, despite the fat that it is an important surrogate fuel for the study of methyl ester-based bio-diesels. A detailed mehanism has been developed at LLNL that predits the behavior of methyl deanoate oxidation [45] BEXPERIMENTAL SETUP For these ignition delay time measurements, the aerosol shok tube setup and supporting diagnostis are shown in Figure 4.1. As in all experiments a series of fastresponse pressure sensors are used for measuring the inident shok speed. A mid-ir (3.39 µm) and a visible ( nm) laser is used for measuring fuel loading. A Kistler TM 601B piezoeletri transduer loated near the endwall (3 m) is used to 63

81 measure the ignition delay time. CH* (at 431 nm) and OH* (at 306 nm) are both deteted through sidewall windows loated near the endwall. Figure 4.1: Shemati of the aerosol shok tube with pressure and laserr diagnostis. The pressure sensors are used for shok speed measurement and ignition delayy time determination, the mid-ir HeNe laser is used for absorption-based fuel measurements, the visible laser diode is used for droplet sattering measurements, and the emission measurement is used to measure ignitionn delay time. The ignition delay time an be defined by the time between the arrival of the refleted shok wave at the pressure sensor (loated 3 m from the endwall) and the time at whih the pressure begins its rapid rise assoiated with ignition. This definition of ignition delay time exhibited the most onsisteny and proved to be the most reliable for these experiments. Ignition delay times determined from laser absorption measurements (at 3.39 µm) of fuel onsumption and emission measurements of CH* and OH*, while onsistent with the pressure measurements, exhibited larger satter and were used only to onfirm the values derived from the pressure measurements. 64

82 4.4 RESUL 38BR TS An example of a typial experiment is shown below in Figure 4.2. The top frame (a) shows 670 nm laser extintion (1), pressure (2), andd CH* emission (3) all loated 3 m from the endwall. The bottom frame (b) shows 3.39 µm laser absorption ( ombined with Mie sattering from droplets when droplets are present) by fuel (4), and 660 nm laser extintion (5), both loated 5 m for the end wall. a b Figure 4.2: Example of an ignition delay time measurement. This example was done with a DF-2 (CI 43) / 21% O 2 / argon mixture with = 0.48, T 5 =1197K, and P 5 =7.21 atm. The diagnostis in the upper frame (a) are loated 3m from the endwall, and in the lower frame (b) are 5m from the end wall: (1) Mie sattering extintion, (2) pressure, (3) CH* emission, (4) fuel absorption, and (5) Mie sattering extintion. Complete evaporation of the aerosol droplets behind the inident shok wave is evident from the 660 and 670 nm laser extintion traes,, whih show zero extintion (at - 20 µs in the upper frame (a) and at -60 µs in the lower frame (b)) before the arrival of the 65

83 refleted shok wave. This wavelength range was seleted as it does not overlap with any absorption feature of the fuel (DF-2 in this ase), and responds only to Mie sattering by droplets. Spikes in the laser signals at 0 µs (a) and 100 µs (b) are a result of beam steering off the detetors from the transiting shok wave. The 3.39 µm laser absorption measurement of fuel (4) in the lower frame (b) verifies that the fuel onentration that passes the window at 5m is uniform. This is evident from the onstant absorbane by fuel-vapor seen from -60 to 100 µs after the omplete evaporation of the aerosol droplets. The test gas mixture that passes the 5 m observation station will stagnate behind the refleted shok wave at the 3 m observation station. Thus, the fuel onentration of the fluid element that stagnates at 3 m from the end wall behind the refleted shok wave will also be uniform. This uniformity is very important to making quality measurements using an aerosol shok tube [42]. The pressure profile (2) in the upper frame (a) shows lassi shok tube behavior. At times before -80 µs, the pressure is given by the pre-shok fill pressure; from -80 to 0 µs, the pressure behind the inident shok wave is reorded; after 0 µs, the refleted shok region pressure is reorded. At 280 µs the rapid rise in pressure signifies ignition. This is oinident with the rise in CH* emission (3) and slightly delayed (at 300 µs at the 5 m loation in the lower frame (b)) by omplete onsumption of fuel. At 5 m the ignition delay time that is indiated by the fuel onentration going to zero (around 200 µs) is shorter than that given by pressure and CH* emission at 2 m. This is likely due to an aeleration of the reation rate upstream due to the strong energy release near the endwall. This ignition delay time has an unertainty assoiated with many aspets of the measurement. The most signifiant ontributor to the unertainty of ignition delay measurements are the unertainties assoiated with the temperature. Beause ignition is so sensitive to temperature any hange in temperature results in a large hange in ignition delay time. The temperature is alulated through AEROFROSH as shown in Chapter 2, and the measurement whih ontributes most substantially to the error in the onditions is the typially the shok speed. We an measure the shok speed with an error of <0.5%. 66

84 This translates to an error in T 5 of <1%. An error of 1% in temperature orresponds to an error in ignition delay of about 10-15%. There are many otherr soures of error like pressure measurement, fuel ross setion, and non-uniformity; however, the unertainty due to the veloity measurement is typially the most severe. In some ases at the longer ignition delays >1 ms, the non-ideal faility effets ausee a temperature error that exeeds 1%, so in these ases the most signifiant error is due to these nonn ideal faility effets. (See Appendix C for a more detailed disussion on the faility effet, namely a +dp/dt) BN-D DECANE Beause n-deane has a higher vapor pressure att room temperature only a sparse aerosol is needed to reate a stoihiometri mixture. The fration of fuel in the final evaporated mixture that omes from the liquid droplets as ompared to the amount in the saturated bath gas is small; n-deane represents one of the most volatile fuels that an be used with the aerosol shok tube methodology. Figure 4.3 shows representative data taken with the aerosol shok tube at 5 atm. Figure 4.3: n-deane/air ignition delay times. Data onditions ranging over P= atm and = are saled to P=5 atm, phi=1.0 using the Olhanski and Burat orrelation [61] (blue diamonds: data; blue line: orrelation; blak lines: best fit to Shen et al. [48]). 67

85 Also shown are the Olhanski and Burat measurements of [61] high-temperature ignition delay times for n-deane in 23% O 2 in argon, obtained in a heated shok tube. Their ignition delay time orrelation an be rearranged to represent the dependene on the variables P and for onstant O 2 mole fration as ~.... Also shown in Figure 4.3 are the reent data at higher pressures (nominal values of 11 and 40 atm) published by Shen et al. in 2009 [48]. The expeted trend with pressure is apparent, and these data exhibit negative temperature oeffiient behavior at high pressures and low temperatures. Both of the previous studies utilized a heated shok tube (100 C in [61]; up to 160 C in [48]). Beause n-deane has a relatively high vapor pressure, only moderate shok tube temperatures are required to generate suffiient fuel loading for equivalene ratios of unity. As a result, pre-test deomposition or oxidation is likely not a problem and thus the agreement between the previous n-deane data and the aerosol shok tube data at 5 atm provides additional onfidene in the aerosol shok tube methodology. As less volatile fuels are used, the heated shok tube must operate at higher temperatures and beomes more suseptible to pre-test deomposition, whereas the aerosol shok tube does not suffer from this unertainty BN-DODECANE Beause of the importane of n-dodeane as a jet fuel surrogate, several ignition delay time studies in heated shok tubes exist. Figure 4.4a presents a omparison of the urrent study with the heated-tube measurements of Shen et al [48] and Vasu et al [52]. 68

86 Figure 4.4: n-dodeane/air and n-dodeane/21% ( =0.5) at various pressures. a O 2 /Ar ignition delay times for fuel-lean mixtures b This omparison of the n-dodeane ignition delayy times measured in heated shok tubes and the aerosol shok tube also shows exellent agreement. A pressure saling of P -0.89±0.09 is found for the air experiments (for all data exept the Shen et al. 40 atm and Vasu et al. 20 atm data at the lowest temperatures), and a saling of P -0.86±0.20 is found for the 21%O 2 /Ar experiments. The error in the fit oeffiient is the standard deviation of a least squares regression. Figure 4.4b showss experiments ( = ) saled to =0.5 as -1.51± ±0.14. The Westbrook et al. (LLNL) [43] and the Wang et al. (JetSurF 2.0) [63] models predit ignition delay time differenes of less than a few perent when N 2 and Ar are interhanged as diluents, whih is onsistent with experiments. The long test times (>10 ms) in Figure 4.4b are no longer than 8 ms, but are saled by equivalene ratio to appear longer. See Appendix C for a disussion on faility limited test times. 69

87 Figure 4.5: Variation of n-dodeane/21% O 2 /Ar ignition n delay time with equivalene ratio saled to 6.0 atm ( atm). In the ase of 21% O 2 / argon, the urrent data shown in Figure 4.5 over equivalene ratios ( ) of 0.05 to The data vary with a saling whih is not a simple power law; however, between 0.31 and 1.31 the dataa roughly follow a saling of -1.51±0.14, whih was used to sale all the data inn Figure 4.4b to = The dependene of ignition delay time on equivalene ratio an be seen in Figure 4.6. At high temperatures there are major differenes between the dataa and the preditions of the detailed mehanisms of Westbrook et al. (LLNL) [43] and that of Wang et al. (JetSurF 2.0) [63]. At low temperatures the data and preditions give the same trend, but deviate at high equivalene ratio. 70

88 a b Figure 4.6: Equivalenee ratio dependene of the ignition delay times for n-dodeane at a) 1250K and b) 1000K, both at P=6.0 atm. The dependene of ignition delay time on equivalene ratio when the oxygen mole fration is fixed is a diret result of hanging the fuel mole fration. At all temperatures studied, the measured ignition delay times are redued for higher equivalene ratios. The models apture this trend at lower temperatures (i.e K), however, at higher temperatures (i.e. 1250K), the models predit that higher fuel onentrations result in a longer ignition delay time. This behavior has also been observed in experiments with smaller fuels (e.g. n-butane) [58]. This behavior may be a result of exessive radial savenging by fuel deomposition produt moleules in the 71

89 model (e.g. OH+alkenes) or redued prodution of radials at higher temperatures (e.g. slower deomposition pathways that generate H-atoms) BN-HEXADECANE n-hexadeane has a very low vapor pressure and experiments by Penyazkov et al. [14] using this fuel represent the urrent limit for existing heated shok tube data. The vapor pressure at room temperature is only 1.4 mtorr, while a 100 torr initial pressure and a stoihiometri mixture in air would require a fuel partial pressure of 0.8 torr (P sat = 109C). In effet, this requires that heated shok tube experiments be made at very high initial temperature. Figure 4.7a- shows a omparison between the urrent data and that of Penyazkov et al [14] (heated shok tube 100C). Figure 4.7a presents the trend of ignition delay time with varying oxygen onentration (for the 1% and 4% O 2 in argon data, τ ~ X -0.54±0.08 O2 ). However, one would expet that the variation would be larger between the aerosol shok tube data (1% and 4% O 2 in argon) and the heated shok tube data (21% O 2 in N 2 ) given the 5-fold inrease in oxygen onentration. Simulations of these experiments using the LLNL C16 mehanism [43] (whih demonstrates a X O2 saling) suggest that the ignition delay times for the 21% O 2 in N 2 experiments would be expeted to be approximately one third of the 4% O 2 in argon experiments, whereas a redution of only ~33% is seen in the data. Fuel deomposition or oxidation to more stable omponents is a possible explanation of this differene as the heated shok tube in the study was kept at 100C and the mixing tank kept at 200C, with mixtures kept in the mixing tank for between 3 and 4 hours. Figure 4.7b and show the variation of n-hexadeane ignition delay times with pressure and equivalene ratio. For fixed oxygen onentration, the measured ignition delay times sale as P -0.73±0.09, and at 4 atm, sale as 0.82±0.13d. Both of these are onsistent with the LLNL C16 mehanism preditions of P and 0.55, though the model saling with equivalene ratio is less satisfatory. 72

90 a b Figure 4.7: (a) Comparison of n-hexadeane ignitionn delay times at 4 atm and equivalene ratio of 1.0 for various oxygen onentrations. (b) Ignition delay time variation with pressure for stoihiometri n- hexadeane / 4% O 2 / Ar mixtures. () Ignition delay time variation with equivalene ratio at 4 atm for stoihiometri n-hexadeane / 4% O 2 / Ar mixtures. 73

91 BMETHYL DECANOATE Methyl deanoatee also has a low vapor pressure (37 mtorr) at room temperature, making it an ideal andidate for measurementss in the aerosol shok tube. Figure 4.8 presents measured ignition delay times for methyl deanoate and simulations using the Herbinet et al. model [46]. These initial experiments were performed at a low equivalene ratio. The experiments have longer ignitionn delay times than the LLNL model simulations [46], by a fator of 2.5, however, the measured ativation energy (44.3±5 kal/mole) is in lose agreement with the LLNL model value (43.6 kal/ /mole). The small pressure range of the experiments ( atm) preluded determining a pressure dependene and so we were unable to onfirm the strong variation with pressuree (τ ~ P ) predited by the model. Over the range of lean equivalene ratios studied, = , the ignition delay times saled as -1.2±0.1. This is in ontrast to the model s predited saling of -4 at these extreme equivalene ratios. Figure 4.8: Ignition delay times of methyl deanoate / 21% O 2 / Ar mixtures at = 0.1 and P=8atm and the ignition delay times from the LLNL mehanism by Herbinet et al. [46]. 74

92 BDIE ESEL Diesel fuel ontains very heavy omponents andd is thus an ideal fuel to be tested in the aerosol shok tube. We have tested a variety off diesel fuelss with varying etane indies and oxygen onentrations. Representative ignition delay times are shown in Figure 4..9 Figure 4.9: Diesel ignition delay times for three different diesel fuels ompared with simulations using the LLNL mehanisms for various surrogate mixtures. Mixtures havee an oxygen onentration of 21% with P=6atm and =0.5. Colored, square symbols indiate data taken with aerosol shok tube; blak irles represent data taken in a heated shok tube by Penyazkov et al. [55]. This figure shows data from 3 different diesel fuel mixtures.. The red data points are from a diesel with a etane index of 43 alulated (ASTM D976) using its density of 0.86 g/ (at 15 C) and a distillation D86 at 50% of C, and was provided to our laboratory from the US Army Researh Offie. The blue data points are from a European ommerial diesel sample, whihh had a higher estimated etane index of

93 These data exhibit the shortest ignition delay times as generally expeted. These measurements were both made in the aerosol shok tube. The blak data points are from measurements by Penyazkov et al. [55] using a heated shok tube (heated to between C). Also shown on this plot are simulations by Westbrook et al. (LLNL) [personal ommuniation 2010] using three different surrogate blends. The surrogate represented by a dotted line, a three-omponent mixture, is based on smaller fuel omponents, x mole % deane, y mole % toluene, and z mole % iso-otane. The theoretial etane index of this surrogate is 55, it overpredits the ignition delay times of the CI=55 European diesel fuel. The other two surrogates (solid red and blue lines) onsidered employed larger moleules that more losely math the moleular weight of diesel fuel, and are based on primary referene fuel, PRF, omponents. These are mixtures of x mole % hexadeane (etane) and y mole % iso-etane for CI=43, and x mole % hexadeane (etane) and y mole % iso-etane for CI=55. Differenes are seen in the experimental results at high temperatures, but not in the simulations using the two different etane indexes. Also see Appendix C Figure C.0.3 for a disussion about the faility effets aused by a slight inrease in temperature and pressure. The US diesel was also tested over a range of pressures and equivalene ratios. Figure 4.10a and b show how these variations effet the ignition delay time. 76

94 a b Figure 4.10: Variation of ignition delay time with pressure ( =0.5) and equivalene ratio (P=6atm) for mixtures onsisting of DF-2 (CI 43) / 21% O 2 / Ar. As pressure inreases the ignition delay time dereases following a saling of P -0.82±0.15. A derease in ignition delay is also apparent if we inrease the equivalene ratio (while holding oxygen onentration onstant) following a saling of -0.70±0.16. An overall orrelation for this fuel with 21% O 2 in argon is given by: 2.64x10 P at tm / / e 77

95 To show variation with the fuel blend, ignition delay times for two other diesel fuels were measured at lower oxygen onentration. The results are shown below in Figure Figure 4.11: Variation in ignition delay time with aromati ontent of diesel fuel. Oxygen onentrations of 4% were used with Ar as the diluent. P=6atm and =0.5. The two diesel fuel blend measurements were ompared to simulations of mixtures of n-hexadeane and iso-etane delay times due to the lower oxygen onentrationn (4% in argon). The two diesel blends were hosen to braket the large (CN 42, 46) using LLNL mehanisms [43, 44]. The data in Figure 11 have inreased ignition variation in aromati ontent allowed by the USS EPA. The higher aromati diesel has a general omposition of: saturates 44. 2%, aromatis 38.8%, and olefins 17%; whereas the lower aromati diesel has a omposition: saturates 81%, aromatis 16.2%, and olefins 2.7%. No signifiant differene was seen in the PRF surrogate (using only etanee and iso-etane as omponents) model for the two etane index ases. The simulations approximately apture the lower aromati fration fuel ignition delay times, but the higher aromati fuel shows a larger disrepany,, implying that this simple mixture is not suffiient to apture the ignition behavior. The agreement may improve by using a mixture whih suffiiently aptures the aromati hemistry that is affeting the ignition 78

96 delay times. In order to better haraterize this differene, we have measured several speies time-histories for many of the low oxygen onentration experiments and these results (still in preparation for later publiation) may help to eluidate the disrepany BCONCLUSIONS This hapter demonstrates the apabilities of aerosol shok tubes to investigate ignition delay times for many low-vapor-pressure fuels. The data obtained provide useful kineti targets for testing the overall behavior of detailed mehanisms for pure and pratial fuels. Several important onlusions an be derived from these new data. Perhaps the most important is that n-dodeane ignition delay times for mixtures with 21% O 2 exhibit a negative power law dependene with equivalene ratio at high temperatures, whereas the LLNL [43] and JetSurF [45] models both predit a positive exponent. We believe that urrent dodeane mehanisms may require more than just minor modifiation to apture this behavior aurately. To our knowledge, this study provides the first shok tube ignition delay time data for methyl deanoate and the first aerosol shok tube measurements for n-hexadeane. The n-hexadeane data provide oxygen onentration dependene, pressure dependene and equivalene ratio dependene. The methyl deanoate data provide temperature and equivalene ratio dependenes, and provide targets for the validation of large n-alkane and methyl ester (i.e. bio-diesel) reation mehanisms. This study also provides gas-phase ignition delay times for several types of diesel fuel and the variation of ignition delay time with etane index, pressure, equivalene ratio, and aromati ontent. These data provide a fundamental database for the testing and refinement of diesel surrogate mehanisms. The data and onditions are summarized in a table in Appendix E. 79

97 Chapter 5: 9BAPPLICATION 2 - SPECIES TIME-HISTORIES The first simultaneous multi-speies laser-absorption time-history measurements for OH and C 2 H 4 during the oxidation of n-hexadeane and ommerial diesel fuel (DF- 2) were aquired. The experiments were performed behind refleted shok waves in the seond-generation aerosol shok tube over a temperature range of 1120K to 1373K and a pressure range of 4 to 7 atm. Initial fuel onentrations varied between 150 and 1800 ppm with equivalene ratios between 0.4 and 2, and were determined using 3.39 µm He- Ne laser absorption. OH onentration time-histories were measured using absorption of frequeny-doubled ring-dye laser radiation near nm. Ethylene time-histories were measured using absorption of CO 2 gas-laser radiation near 10.5 µm. Comparisons are given of these speies onentration time-histories with two urrent large n-alkane mehanisms: the LLNL-C-16 mehanism of Westbrook et al. (2008) and the JetSurF C- 12 mehanism of Sirjean et al. (2009). Fair agreement between model and experiment is seen in the peak ethylene yields for both fuels; however, modeled early time-histories of 80

98 OH, an important hain-branhing speies, differ signifiantly from urrent measurements BINTRODUCTION Detailed speies time-history measurements during reation of gas-phase diesel fuel and diesel fuel surrogates are needed to test and refine the large reation mehanisms used to model the ignition of these fuels. The aerosol shok tube is well-suited to fill this need; all that is required is to equip the shok tube with the relevant laser diagnostis. In our laboratory, we have developed sensitive, speies-speifi, and quantitative UV and visible laser absorption diagnostis for an array of important ombustion speies inluding OH, CH 3, CH, NH 2, and NO [64]. Reently, we have extended our suite of laser absorption diagnostis into the infrared. Using ommerially available IR gas and diode lasers we have developed quantitative measurement apabilities for C 2 H 4, H 2 O, CO 2, CH 4 and seleted n-alkanes [65-67]. Use of these laser absorption diagnostis in shok tubes has allowed the generation of multi-speies time-histories under nearonstant-volume onditions intended as targets for hemial kineti modeling. To eluidate the inner workings of the large reation mehanisms desribing diesel and diesel-surrogate ignition hemistry, we have applied this multi-speies strategy to two important transient speies that appear during diesel and n-hexadeane oxidation: OH, and C 2 H 4. Time-history measurements of OH are extremely useful in quantifying ignition proesses, as they provide ritial information about the radial-pool population. OH profiles in these systems typially show an immediate rapid rise (that is dependent on fuel deomposition kinetis) to a relatively long-lived pre-ignition plateau (that is dependent on the slower oxidation kinetis of the stable alkene intermediate speies), and a final strong exponential rise during ignition. Time-history measurements of ethylene, C 2 H 4, provide quantitative information on the main high-temperature deomposition produt and pathways of alkanes. Other groups have examined multiple speies during hexadeane and diesel ombustion in a jet-stirred reator (JSR) [68-70], but these studies 81

99 do not provide the detailed information on the short, transient time sales available in shok tubes. Here we present measurements of OH and C 2 H 4 during the oxidation of diesel fuel (DF-2) and a single-omponenare given of these speies onentration time-histories with two urrent large n-alkane mehanisms: the LLNL-C-16 mehanism of Westbrook et al. (2008) [43] large-alkanee surrogate for diesel fuel, n-hexadeane. Comparisons and the JetSurF C-12 mehanism of Sirjean et al. (2009) [71] BEX XPERIMENTAL SETUP 84BAEROSOL SHOCK TUBE The experiments were performed in thee seond-generation aerosol shok tube (AST II). This faility is shown shematially in Figure 5.1. Speies onentration timeshok history measurements are then made using laser absorption behind the refleted wave throughh sidewall windows loated at 4 m from the endwall. In Figure 5.1, the refleted shok has propagated nearly all the way through the aerosol mixture (labeled C). Figure 5.1: Shemati of aerosol shok tube (AST II) setup. The different regions in the tube are (from left to right): A - driver gas, B - inident-shok heated driven gas with no fuel, C - inident-shok heated test mixture with fuel, and D - refleted-shok heated test mixture. 82

100 BLASER ABSORPTION Three CW laser systems are used in these experiments: an IR He-Ne laser to measure initial fuel onentration; an IR CO 2 gas laser to measure ethylene; and a UV ring dye laser to measure OH onentration. Initial fuel onentration measurements were by laser absorption at µm using a low-noise IR Helium-Neon laser (Jodon TM model HN-10G-IR). Signifiant interferene absorption by fuel deomposition produts that are formed during hightemperature oxidation prevents this single-wavelength absorption signal from being used to monitor the full time-histories of the parent fuel. Ethylene was measured at two wavelengths using tunable CO 2 laser absorption. The wavelengths that were used for this experiment were µm for on-line C 2 H 4 measurements (strongest absorption by C 2 H 4 ) and µm for off-line C 2 H 4 measurements (weak absorption by C 2 H 4 ). The laser was tuned to the speifi CO 2 transition (P(14) for on-line and P(28) for off-line) Potential lose-lying transitions of the CO 2 laser were suppressed by first passing the laser beam through a monohromator. A ommon-mode rejetion sheme was used to redue laser noise (for all three laser systems), employing both referene (I 0 ) and transmitted (I) beams. The transmitted CO 2 laser beam passed through the shok tube, through a narrow band-pass filter, and was foused onto a liquid nitrogen ooled MCT detetor (Fermionis TM PV ). The detetion sheme is similar to that used by Pilla et al. [65]. Speies onentration for ethylene (and for fuel and OH) were determined using the Beer-Lambert relation (in a slightly different form than in Equation 2.2): -ln(i/i 0 ) = -k ν PXL, where k ν is the absorption oeffiient [atm -1 m -1 ], P is the total pressure [atm], X is the absorbing speies mole fration, and L is the pathlength, 10 m in the urrent experiments. Separate refleted shok tube and FTIR experiments, to determine the absorption oeffiients for ethylene [65] and fuel, were also performed. Absorption oeffiients for OH are wellestablished [72]. Two wavelengths were used to separate the C 2 H 4 absorption from the interfering absorption of other alkene speies formed during ignition. Near 10.5 µm, the seondary 83

101 (interfering) absorbers are primarily larger olefins, suh as C 3 H 6, C 4 H 8, and the ombustion produts H 2 O and CO 2. (Signifiant absorption from the ombustion produts ours only after ignition, when ethylene onentrations are negligible, and is not inluded in the following analysis.) The wavelengths of the absorption peaks of these olefins are signifiantly removed from, and do not overlap with, the C 2 H 4 features at µm and µm, the wavelengths of the P(14) and P(28) transitions, respetively. At these two wavelengths, absorption from the far wings of these olefins bands is assumed to be approximately onstant with wavelength. With this assumption and aurate measurements of the absorption oeffiient of C 2 H 4 at the two wavelengths, we are able to separate the absorption ontributions of the ethylene and interfering absorbers, and uniquely determine the C 2 H 4 onentration time history. OH onentration time-history was measured using laser absorption of the R 1 (5) transition in the OH A 2 + -X 2 π(0,0) band at nm. This UV wavelength was generated using a 532 nm laser (Coherent Verdi) to pump a ring-dye laser (Spetra Physis 380) that was intra-avity frequeny-doubled using a temperature-tuned AD*A rystal. The laser set up was similar to that used by Herbon et al. [72] and others in our laboratory. A similar two-wavelength tehnique was used to separate the OH absorption from interferene absorption seen near 306 nm. In this wavelength region there is broadband absorption, likely from onjugated olefins (e.g. 1,3-butadiene). Off-line measurements ( m -1 ) between the R 1 (5) and the R 1 (7) transitions have effetively no absorption by OH, and any interferene absorption (assumed to be onstant with small wavelength hange) an be diretly subtrated from the on-line measurements to give a diret measure of the OH onentration time-history. As in the ase of ethylene, two shok wave experiments are required at near-idential onditions to aquire the measurements at two wavelengths and effet the subtration. Representative data using this method are shown in Figure 2. High SNR using the UV OH laser absorption diagnosti enables aurate subtration of the two signals and results in minimum detetion limits for OH of 1-10 ppm. 84

102 Strong interferene absorption in the UV near 306 nm was seen in all of the diesel experiments as indiated in Figure 5.2. However, the hexadeane experiments showed no signifiant UV interferene absorption. Related behavior was seen by Davidson et al. [73], where interferene absorption was observed duringg iso-otane oxidation, but not in measurements during oxidation of normal alkanes [74]. Figure 5.2: Diesel OH on-line and off-line laser absorption near nm and the differene absorbane signal attributable solely to OH during diesel oxidation. The pre-ignition plateau orresponds to around 10 ppm of OH. Initial refleted shok onditions: ONLINE: 1198 K, 6.69 atm, 1228 ppm diesel; OFFLINE: 1193 K, atm, 1197 ppm diesel both in 4% O 2 / argon. 5.3 RESUL 42BR TS AND DISCUSSION Presented here are laser absorption measurement ts of OH and C 2 H 4 onentration time-histories during the high temperature oxidation of n-hexadeane and diesel fuel. These measurement ts are also ompared with model simulations of n-hexadeane oxidation using the LLNL C-16/Westbrook et al. (2008) mehanism [43], and model simulations of n-dodeane and n-heptane oxidation using the JetSurF 1.0/Sirjean et al. 85

103 (2009) mehanism [71]. All alulations were performed using the Reation Design Aurora suite of CHEMKIN-based odes and implemented with a onstant U-V (onstant internal energy - onstant volume) onstraint. n-hexadeane was proured from Sigma-Aldrih (99+%) and degassed only before use. The diesel sample (DF-2) was proured loally and degassed only before use. Analysis by IAC Laboratories San Franiso indiate that for the DF-2 fuel, the C/H mass ratio is 6.26, the volume ratio of aromati:olefin:saturated is 16.2:2.7:81.0, and the etane index is The hemial formula used for equivalene ratio determinations for the diesel fuel was C 12 H N-HEXADECANE OXIDATION Figure 5.3 shows OH and C 2 H 4 time-histories during n-hexadeane oxidation at 1267 K, 6.54 atm, in 1% O 2 /argon with an equivalene ratio =1.21. Exellent signal to noise ratio (SNR) is seen at all times in the C 2 H 4 data; however, the low levels of OH measured (~2 ppm) before ignition result in a smaller SNR. The simulations shown using the Westbrook et al. [43] C-16 mehanism (labeled LLNL model) effetively apture the formation rate and nearly apture the pre-ignition C 2 H 4 plateau value. Similarly, the modeled time of ignition, indiated by the rapid rise in OH onentration mathes the measurement, though the measured OH pre-ignition plateau (at times of 0 to 1000 µs) is approximately twie the modeled value. 86

104 Figure 5.3: OH and C 2 H 4 speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: =1.2, 1267 K, 6.54 atm; initial test gas mixture: 497 ppm C16H 34, 1% O 2 /argon. The onstant UV simulation is based on a LLNL hexadeane mehanismm by Westbrook et al. [43]. Figure 5.4 and Figure 5.5 show how the C 2 H 4 and OH time-histories and the Westbrook et al. C-16 mehanism simulations vary with temperature. The simulations for the lowest and highest temperature examples overpredit the ethylene yields, while the middle temperature (1170 K) simulation is only slightly overpredited. From the overpreditions of the ethylene yields, it an be inferredd that fuel deomposition produt yields of other larger olefins (e.g. propene and butene) are likely underpredited. As well, at these high temperatures, formation times of C 2 H 4 are diretly related to fuel deomposition rates and pathways, and future efforts to math these profiles should result in improved rates for the deomposition and oxidation pathways for this large n-alkane. Also plotted are C-16 mehanism simulations with faility effets inorporated into the simulation. In these experiments, there was a slight pressure inrease behind the refleted shok (dp/dt~0.2 atm/ms) see Appendix C andd Figure C.0.2 for an explanation and a harateristi pressure trae. This pressure risee only signifiantly affeted the lowest temperature ase by shortening the ignition delay. The ethylene yield stayed the same, so the disrepany in ethylene yield is most likelyy not due to faility effets at the lowest temperature and is an indiation of a problem withh the hemial mehanism. 87

105 Figure 5.4: C2H 2 4 speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: = =0.8, 1170 K, 4.60 atm, 326 ppm C 16 H 34, 1% O 2 /argon; =1.2, 1267 K, 6.54 atm, 497 ppm C16H 34, 1% O 2 / argon; and = =1.2, 1333 K, 6.77 atm; 4933 ppm C 16 H 34, 1% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] with onstant UV and aounting for faility dp/ /dt. Figure 5.5: OH speies time-histories during n-hexadeane oxidation. Initial refleted shok wave onditions: = =0.8, 1170 K, 4.60 atm, 326 ppm C 16 H 34, 1% O 2 /argon; =1.2, 1267 K, 6.54 atm, 497 ppm C16H 34, 1% O 2 / argon; and = =1.2, 1333 K, 6.77 atm, 4933 ppm C 16 H 34, 1% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] with onstant UV and aounting for faility dp/ /dt. 88

106 Peak OH onentrations are aurately modeled at the two lower temperatures (1170 and 1267K) simulations, but not for the highest temperature ase. Given that OH onentrations are losely related to the other important small radial speies (H-atoms, O-atoms and HO 2 radials) that ontribute to hain branhing, efforts to refine rates and pathways in the mehanism that improve preditions of OH should improve preditions of all the small radials ontributing to ignition progress. A detailed review of all hainbranhing reations would be required to understand the ause of the disrepany between the model and the measurements DIESEL OXIDATION Figure 5.6 shows OH and C 2 H 4 time-histories during diesel oxidation at 1198 K, 6.69 atm, in 4% O 2 /argon with an equivalene ratio =0.54. As in the ase of n- hexadeane, good SNR is seen in the ethylene trae. Interestingly, improved SNR is seen at early times in the OH trae. Simulations using three single-omponent diesel fuel surrogates, C 7 H 16, C 12 H 26 and C 16 H 34, (using mixtures with idential oxygen mole fration and equivalene ratios) are also shown. While all three of the diesel fuel surrogate simulations give similar ethylene peak yields (and similar OH peak values), none of the simulations losely mathes the ignition delay time (derived from the time to half peak OH onentration). As well, none of the three single-omponent surrogate models aurately predits early time OH onentrations and all three models underestimate the start of the OH plateau by an order of magnitude. Given that this diesel sample is 81% saturated paraffins, it might have been expeted to be modeled relatively well by a single large n-alkane. The approximately 30% variation between the modeled (JetSurf and LLNL models) and experimental ignition delay times is a measure of the goodness-of-fit one an expet using any of the urrent single-omponent diesel surrogates. Post-ignition values of ethylene are expeted to be near zero; however absorption by the ombustion produts H 2 O and CO 2 begins to appear at ignition, and thus the absorption-based C 2 H 4 profile is not shown beyond 1200 µs. 89

107 Figure 5.6: OH and C 2 H 4 speies time-histories during diesel oxidation. Initial refleted shok wave onditions: = =0.54, 1198 K, 6.69 atm; initial test gas mixture: 1228 ppm diesel, 4% O 2 /argon. Constant UV simulationss for single-omponent surrogate models shown are: n-heptane and n-dodeane whih use the JetSurf mehanism and n-hexadeane whih uses the LLNL mehanism. Figure 5.7 and Figure 5.8 present variations in the OH and C 2 H 4 time-history measurement ts as a funtion of temperature for diesel oxidation. In both the OH and the C 2 H 4 time-histories, modeled values of peak yields (using the n-hexadeane single- values at all temperatures s. Time sales in the simulations for formation of OH and C 2 H 4 omponent surrogate) are relatively lose to (but onsistently exeed) the measured typially vary by a fator of two from the experiment. Given the urrent relatively onfident state of knowledge about large n-alkane pyrolysis and oxidation hemistry (albeit tempered by the omparisons shown earlier for n-hexadeane) ), the differenes seen between the single-omponent surrogate modeling and the diesel experiments indiates thatt a multi-omponent surrogate for diesel will be needed to apture the finer details of these experiments. In partiular, interfering absorption near nm that indiates the presene of UV absorbing speiess (i.e. onjugated olefins) during diesel oxidation hemistry, but not during n-hexadeane oxidation, point to a need to inlude these larger, more hemially omplex, intermediate speies in diesel modeling. 90

108 Figure 5.7: C 2 H 4 speies time-histories during diesel oxidation. Initial refleted shok wave onditions: =0.8, 1119 K, 4.63 atm, 1761 ppm diesel, 4% O 2 /argon; =0.5, 1198 K, 6.69 atm, 1228 ppm diesel, 4% O 2 /argon; and = K, 6.33 atm, 1380 ppm diesel, 4% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] for n-hexadeane as the fuel with onstant UV and aounting for faility dp/dt. Figure 5.8: OH speiess time-historiess during diesel oxidation. Initial refleted shok wave onditions: =0.8, 1119 K, 4.63 atm, 1761 ppm diesel, 4% O 2 /argon; =0.5, 1198 K, 6.69 atm, 1228 ppm diesel, 4% O 2 /argon; and = K, 6.33 atm, 1380 ppm diesel, 4% O 2 /argon. Modeled using the LLNL model by Westbrook et al. [43] for n-hexadeane as the fuel with onstant UV and aounting for faility dp/dt. 91

109 The inreased noise in the C 2 H 4 and OH profiles in Figure 5.7 and Figure 5.8 respetively is also a result of a redution in SNR due to subtration of absorption from interfering speies. Also the inlusion of the faility dp/dt shortened the ignition delay in the oldest experiment BCONCLUSIONS The first multi-speies time-history measurements for high-temperature n- hexadeane and diesel oxidation are presented. Measured time-histories for OH and C 2 H 4 are ompared with single-omponent diesel surrogate models. Agreement in ignition delay times and speies onentrations typially are within a fator of two at all temperatures. However, early-time preditions of OH mole fration, indiative of the ability of these mehanisms to apture the important hain-branhing speies onentrations, vary widely from the experiments. These measurements provide quantitative kineti targets for the testing and refinement of n-hexadeane and diesel surrogate models. Further work is planned to extend speies time-history measurements for these fuels to inlude simple alkanes (i.e., methane and ethane), larger alkenes partiularly propene and butene, and onjugated alkenes (e.g., 1,3-butadiene). 92

110 Chapter 6: 10BCONCLUSIONS 6.1 4BFACILITY AND METHOD A new faility and method of studying the hemial kinetis of low-vaporpressure fuels has been developed. The aerosol shok tube was modified to a seondgeneration aerosol shok tube, whih displayed signifiant improvement in aerosol spatial uniformity. Previously the aerosol shok tube relied on a ontinuous flow with turbulent mixing to fill the shok tube. The improved faility uses a non-ontinuous, laminar plug flow. This required the onstrution of a mixing tank, whih is onneted diretly to the endwall, a speially designed endwall gate valve, driven gate valve, and a dump tank. The aerosol is reated in the mixing tank, whih is apable of aerosol loadings up to 5 times higher than those possible with the first-generation aerosol shok tube. The aerosol then flows into the test setion of the driven setion of the tube between the gate valves. This is done by gently pulling the aerosol through expansion into the dump tank. This method of filling produed signifiant improvements in the spatial uniformity of the aerosol. This uniformity was shown to impat the reproduibility of the hemial kineti measurements that were made. 93

111 6.2 45BIGNITION DELAY TIMES Using this new faility and tehnique, measurements of ignition delay times for several low-vapor-pressure fuels were made inluding n-deane. n-deane has been studied quite extensively in the past using the heated shok tube method, whih for this fuel we do not worry about premature deomposition. We see good agreement between the heated shok tube and aerosol shok tube measurements. We then measured even lower vapor pressure fuels inluding n-dodeane, n-hexadeane, and methyl deanoate and produed some of the first measurements of ignition delay time for these fuels. We also applied the aerosol shok tube method to the measurement of ignition delay times for real multi-omponent fuels. Ignition delay times for a low volatility jet fuel known as JP-7 were determined using this method [42]. We also measured the ignition delay time of several different bathes of diesel fuel from around the world. We found that etane number affets the ignition delay at high temperatures. Also, that aromati ontent affets the ignition delay as well. This work has also brought attention to the fat that there needs to be better surrogate mixtures and mehanisms to simulate the ombustion of diesel fuel. These high quality ignition delay times, whih are unaffeted by partial distillation and evaporation, provide aurate benhmark validation targets for the development of suh surrogate mehanisms BSPECIES TIME-HISTORIES To improve these validation targets we drew upon our laboratory s ability to measure onentrations of important ombustion speies, and applied unique laser absorption diagnostis to the study of low-vapor-pressure fuels in the aerosol shok tube. We measured OH and C 2 H 4 during the oxidation of n-hexadeane and diesel fuel. To measure OH we used a ppm-sensitive UV absorption diagnosti. For C 2 H 4 we used a mid-ir CO 2 laser. We measured the OH and C 2 H 4 time-histories for a range of temperatures for both fuels. The results for n-hexadeane showed agreement with a mehanism developed at LLNL by Westbrook and oworkers in some ases. For diesel 94

112 fuel we ompared the measurements to a primary referene fuel surrogate mehanism also developed at LLNL, and the agreement was not as good. This is understandable beause the primary referene fuel is not meant to apture the details of the oxidation, only the ignition delay BFUTURE WORK The aerosol shok tube an be applied to many engineering and siene problems and through this work we have found many alternative appliations for this unique faility. For example, others in our laboratory have used it to study low-vapor-pressure pyrolysis, bio-aerosol-shok interations, and nano-aluminum-slurry ombustion. These suessful appliations demonstrate the wide versatility of this faility and beg the question of what else an be done with this faility BNEGATIVE TEMPERATURE COEFFICIENT Many reent studies for smaller fuel moleules have been foused on the negative temperature oeffiient (NTC) behavior. This is where a link in the fuel deomposition slows the overall reation with inreasing temperature for ertain onditions. This ours typially at lower temperatures and higher pressures. Coinidentally this ours at the same onditions where most diesel engines operate. Beause this behavior is highly dependent on the fuel deomposition pathways it is very important that as we extend the experimental database to large fuel moleules with the aerosol shok tube that we inlude the behavior of these fuels in the NTC regime. Muh of this work on the aerosol shok tube was foused on making sure that this new method and faility worked at extending the range of fuels that an be studied using shok tubes. The future of this faility should inlude the measurement of NTC behavior and that will require applying new shok tube tehniques suh as driver extensions and driver inserts pioneered in our laboratory. 95

113 BIOFUELS As shown in Figure 1.1 bio-diesel has one of the highest distillation urves of these pratial fuels. This makes it an ideal andidate for the aerosol shok tube. Also beause of pressure to redue dependene on foreign soures of fuel and to limit the arbon footprint that petroleum fuels ause, biofuels are a promising future soure of energy. In this work, I presented measurements of methyl deanoate ignition delay time. This fuel however is only a smaller version of the moleules that are prevalent in biodiesel blends. The seond-generation aerosol shok tube is fully apable of being used for the large hain methyl esters that are found in bio-diesel (e.g. methyl palmitate, methyl stearate, methyl oleate, methyl linoleate, and methyl linolenate). 96

114 1BAPPENDIX A 48BA.1 LIQUID FILMS Any laser measurement in the aerosol shok tube is onfounded by the fat that we have droplets and highly ondensable fuels. The fuel is slightly elevated in temperature (a degree or two above room temperature) due to the energy involved in nebulization. Then it flows into the shok tube. Beause the aerosol has a large surfae area, the gas surrounding the droplets an be reasonably assumed to have the saturation vapor pressure of the fuel at the temperature of the droplets. Every surfae that this mixture omes in ontat with inluding the windows, whih are a few degrees ooler than the gas, will ondense fuel. As I referred to earlier heating the windows is needed to avoid window fouling. To avoid this ondensation, a few degrees of heating would suffie. The problem worsens though as we perform a shok experiment. After the passage of the inident shok the temperature is instantly raised to somewhere around 600K. At this temperature the saturation vapor pressure is several orders of magnitude above the value at room temperature, and so the aerosol evaporates until there is no liquid fuel left. The fuel vapor partial pressure never reahes the saturation value for the temperature of the bulk gas. However, near the edges of the shok tube there is a boundary layer where the temperature rapidly dereases to room temperature at the shok tube wall, so there exist large gradients in both temperature and fuel onentration. These gradients produe large diffusive fluxes of heat and fuel. The temperature of the wall only rises slightly due to the fat that the shok experiment is so short and the heat ondution of the metal is very high so the rise in temperature is spread quikly into the bulk of the material. The fuel that makes it through the boundary layer is met by, relatively speaking, a very old wall, and so the moleules stik and ondense. This is 97

115 very short so not muh fuel is built up, but thee measurements an be miron thiknesses of a liquid fuel film. affeted by even Figure A.0.1: Examples of shok veloity attenuation for various window heating onfigurations. that the old heater raised the temperature in region 1 aelerating the shok near the endwall. Note The solution to this problem is to heat the windows to temperatures suh that the fuel moleules will not stik to the window, but stay in the gas phase. The temperature where this beomes an effetive method is around 70C. This also beomes a problem however, beause it reates a naturally onvetive flow inside the shok tube as well as a longitudinally non-uniform temperature field in the region near the endwall. The nonuniform temperature field has been seen to affett the shok speed (Figure A.0.1). As the shok propagates into regions with higher temperatures the shok aelerates. This is a problem beause our method of measuring shokk speeds with pressure transduers has a limited spatial resolution and relies on extrapolation of a shok speed versus distane urve, so it is very sensitive to hanges in thee behavior of the attenuation. Beause heating reates these ompliations, it was neessary to minimize the area whihh was heated to minimize the disturbane to the initial onditionss of the shok tube. This is failitated by a heating and ooling sheme built into the mount that holds the windows (Figure A.0.2). The results are windows that are apable of being heated to around 100C, while the wall all around the window is maintained at room temperature. 98

116 49BA.2 2 WINDOW DESIGND The mount is made of brass inside of whihh there is a water-ooled avity surrounding the opper windoww plugs thatt are isolated thermally with a erami that would minimize heat transfer, but withstand the high fores involved in a shok and/or detonation wave. There are band heaters around the opper plugs, whih transfer the heat to the windows and then a flow of old water around the plugs immediately removes the heat. The window mount is shown below in Figure A.0.2. Figure A.0.2: New heated window mount prior to installing on shok tube. Copper tubes are wrapped with ring heaters whih ondut heat to windows. The mount is ooled by flowing ool water through inside of mount. 50BA.3 OPERATIONAL CONDITC TIONS The temperature of the window is usually held att the lowest temperaturee that does not allow window ondensation. Reognizing the presene of ondensation eluidates two distint ways that ondensation ours. The extintion of the visible laser normally indiates the presene of droplets and after evaporation the extintion will typially go to zero and stay at zero. If the extintion raises again slowly in region 2 and faster in region 5, then ondensation is happening at distint nuleation sites on the window. The ondensation reates bumps of liquid on the surfae and satters the beam as these bumps 99

117 grow, the sattering inreases. The inrease is faster in region 5 beause the partial pressure of fuel is higher in this region. The other less ommon method of ondensation is where nuleation sites are not apparent and the only indiation of a liquid on the surfae omes from a negative extintion of the non-resonant beam. The extintion dereases monotonially during evaporation, but asymptotes to a value below zero. The only way for this to happen is if the I is larger than the I o (in the formula ln(i/i o )). So in this ase we first need to rule out emission, whih is not usually a problem in region 2. We find that this is the same effet as an anti-refletion oating on a window. What is happening is that the fuel is ondensing in a smooth film over the window. Beause we are required to use speialized windows like sapphire to transmit the mid-ir light there is a large differene in the index of refration of sapphire (1.7) as ompared to that of the gas mixture (1.0) in the shok tube. Therefore a thin film of fuel (1.4) will effetively redue the net effet of Fresnel refletions, whih initially aused attenuation in the signal. 100

118 12BAPPENDIX B 51B.1 SHOCK AEROSOL INTERACTIONS When the shok wave hits an aerosol the liquid droplets ontained in that aerosol rapidly exhange mass, momentum, and energy with their surroundings. This state of non-equilibrium lasts longer than in a typial purelyy gas-phase shok, and so it is informative to understand what is happening in this transient proess behind the shok wave. Figure B.0.1: X-t diagram for shok aerosol interation. Partile breakup ours fastest with partile aeleration ourring more slowly. Evaporation and diffusion our most slowly after whihh the flow is in equilibrium. The measurement loation is seleted suh that all these proesses have ourred to a suffiient extent upon arrival of the refleted shok. 101

119 The proess is shown on an X-t diagram (Figure B.0.1). This details the position of the shoks in the shok tube. We start at the bottom of the graph where the inident shok is on the left travelling right toward the endwall. Behind the inident shok the quikest transfer happens with momentum. As the shok wave passes the droplet it leaves behind a small bow shok whih reflets off the surfae of the droplet. The droplet then begins to aelerate from the momentum transfer of the subsoni gas whih is deelerated by the bow shok. As the droplet moves toward the endwall it is surrounded by the higher veloity gas and is eventually travelling at the speed of the gas surrounding it. This typially takes less than 10 µs for aerosol used in this study (d ~ 4µm). A somewhat slower proess whih happens at the same time as the momentum transfer is the transfer of mass and energy. This proess takes longer to omplete, with evaporation happening in about 100 µs for our 4 µm aerosol. The transfer of energy ours at the surfae of the droplet. The gas surrounding the droplet is flowing and onvetively transferring heat to the droplet for the first 10 µs. When the droplet aelerates to the same speed of the gas the heat is being transferred to the droplet purely ondutively. A gradient in temperature in the gas at the surfae sustains a heat flux, whih is mathed by a smaller gradient in the liquid droplet. As suh the temperature of the droplet is ever inreasing with the oldest point at the enter of the droplet. The hottest point of the droplet is the surfae whih is evaporating fuel to math the loal vapor pressure surrounding it. And as a result the fuel vapor onentration is higher near the surfae of the droplet reating a diffusive flux of fuel away from the droplet. This evaporation proess onsumes the fuel and the droplet diameter shrinks until it is ompletely evaporated. 52B.2 GAS PHASE DIFFUSION After evaporation we have a purely gas phase mixture of a very low-vaporpressure fuel. This is not the end of the non-equilibrium state though. One the mixture is ompletely gas phase, we are left with loalized old spots due to the loalized 102

120 onsumption of the heat of vaporization, as well as loalized fuel rih pokets. These loalized pokets ould be detrimental to a ombustion experiment resulting in temperature and equivaleny averaged behavior. So we must be sure that these gradients are effetively smoothed out prior to the arrival of the refleted shok. This fuel diffusion problem an be approximated as a spherially symmetri diffusion problem. We will over-estimate the diffusion time if we approximate the fuel as initially onentrated in a small sphere of mole fration equal to unity surrounded by pure bath gas out to a radius equal to half the inter-droplet spaing. At this shell we an assume that a no flux or symmetry boundary ondition is appropriate. This problem is very similar to the free diffusion problem with a self-similar solution. The important result is that the time it takes the fuel to diffuse to a radius R is proportional to R 2 /D where D is the diffusion oeffiient for the fuel moleules diffusing into the bathgas. This result tells us that the diffusion time is strongly related to the distane between the droplets. This gives us insight into how we should design our experiment to minimize any small sale non-uniformities. The amount of time these bumps take to smooth out depends on the radius of the symmetry boundary ondition very strongly. This means that we would like to have as small of an inter-droplet spaing as possible. In setting up an experiment the inter-droplet spaing is set by the equivalene ratio and fuel loading that we want to ahieve for our experiment. But we do have ontrol over the droplet diameter, whih for reasons mentioned earlier (setion 2.2), we noted that it would be advantageous to make the droplet diameter small to trak the flow. As for diffusion times, this is also the ase. If we make the droplet smaller, while keeping the fuel onentration onstant, we need more droplets per volume and hene a smaller interdroplet spaing. We an insert estimates for the values and onditions we are hoping to ahieve and find that the diffusion time is around 100 µs using this rudimentary analysis. If we model this numerially we get similar answers. Below is a plot of the fuel vapor onentration evolving with time for an example ase (Figure B.0.2). Note that the inter-droplet spaing for a C V = 15 ppmv and d = 4.5 µm is about 400 µm. 103

121 Figure B.0.2: Evolution of gas phase fuel onentration. Assuming spherial symmetry and that the droplet instantaneously evaporates. In this ase, it takes only 15 µs forr the fuel onentration to be 95% uniform. If we perform these alulations over a range of liquid volume onentrations (Figure B.0.3) we seee that the diffusion time dereases with inreasing loading. This is beause the inter-droplet spaing dereases. The same thing happens if droplet size dereases, beause theree are more droplets present for a speifi loading. 104

122 Figure B.0.3: Time to ahieve 95% uniformity with varying volume fration of liquid for monodisperse size distribution of various diameters. This analysis presumes a spherially symmetri senario, whih is a simplifiation. When the momentum transfer first ours, the stagnation point on the surfae of the droplet is going to have higher transfer rates breaking the symmetry. Also depending on the shok strength, droplet breakup has been known to our for Webber Numbers above 12. All of this will only serve to enhane mixing and speed up these proesses, so in the purely diffusive spherially symmetri ase we are onsidering is a worst-ase senario. This analysiss is very important, beause it also tells us where to plae our measurement diagnostis in the shok tube. As noted before, the laser measurements are made longitudinally aross the shok tube through two windows in the sidewalls near the endwall of the shok tube. It is advantageous to have this diagnosti as lose to the endwall as possible espeially for high energy release onditions. However due to pratial onstraints the measurement is typially made no less than 2 m from the endwall. The distane the measurement loation is from the endwall diretly determines the time between the arrival of the inident shok and the arrival of the refleted shok. 105

123 This however is the referred to as the laboratory time, and time. The differene is explained in the next setion. is not the same as the fluid 53B.3 L LAB TIME AND PARTICLE TIME The differene between lab time and fluid time is due to the fat that the fluid is passing by the measurement loation ontinuously between the inident shok arrival and the refleted shok arrival. The situation is best understood when looking at the X-t diagram in Figure B.4. Figure B.4: X-t diagram showing the refletion of the inident shok wave at the endwall. Also shown is a partile path for the fluid element that is at the measurement loation in region 5. This partile experiene region 2 onditions for longer than the measurement loation views region 2 resulting in a differene between lab time and partile time. This is a plot of distane on the vertial axis and time on the horizontal axis. The shoks are shown as well as the path of a fluid element. The inident shok starts in the lower left hand orner and travels toward 106 the endwalll and along the way intersets the fluid

124 element and aelerates it toward the endwall. This begins the fluid time. Then the shok passes the measurement loation and the lab time is started. The inident shok then hits the endwall and a refleted shok is reated and travels bak upstream. The fluid element was hosen as the fluid element that intersets the refleted shok at the moment it passes the measurement loation. This ends the fluid time and the lab time simultaneously. This tells us that the time that the fluid element is exposed to region 2 onditions is longer than the time that the measurement loation experienes region 2 onditions. The relationship between the lab time (LT) and fluid time (FT) is Equation B.1 107

125 13BAPPENDIX C 54BC.1 BEHIND THE REFLECTED SHOCK When the refleted shok arrives at the measurement loation it stagnates the flow. At this point the alulation of the shok jump is muh easier beause there is no phase hange. The shok wave is initiated suh that the onditions at this point are suffiient to begin to deompose the fuel. This proess an be observed at the measurement loation optially and typially we have a pressure sensor measuring the pressure. For high energy release onditions pressure rise is a good indiation of ignition. Other things suh as fuel deomposition speies time-histories and emission an be measured as well. 5BC.2 TEST TIME These measurements of onentration and emission are easily modeled if the onditions are not affeted by faility-indued perturbations, the most striking of whih would be the arrival of the expansion fan to the measurement loation. This expansion fan has to travel from the diaphragm to the end of the driver then reflet and travel the entire length of the shok tube before reahing the measurement loation. The effet that this will have is a rapid expansion and ooling of the fluid element along with an aeleration of the fluid away from the endwall. In terms of the ombustion event it will slow and stop any signifiant reations. The arrival of the expansion fan an be delayed by several methods. The omposition of the driver an be altered by lowering the ratio of speifi heat. This will effetively slow the speed of sound of the mixture and slow the expansion wave in the driver mixture. Another method is to make the driver longer. This has been employed in previous studies and a similar shok extension was added to the aerosol shok tube as seen in Figure C

126 Figure C.0.1: Piture of aerosol shok tube driver setion and additional segments for extending driver length (Photo Courtesy of Matt Campbell). 56BC.3 CONTACT SURFACE REFLECTION There is also another effet that an limit the test time even before the arrival of the expansion fan. This ours beause of the existene of a ontat surfae that exists between the driver gas and the driven gas. This interfae travels toward the endwall at the same speed as the gas behind the inident shok wave. The inident shok will arrive at the endwall long before the ontat surfae. As a result the refleted shok as it is travelling bak toward the driver will interset thee ontat surfae. Beause this surfae marks a hange in gas mixtures the speed of sound in eah mixture is different. This differene will ause a slight refletion of the refleted shok wave, whih will travel bak toward the endwall. At the measurement loation a slight pressuree inrease is the result. Along with this inrease is also an inrease in temperature, thus ending our test time. In order to minimize the impat of this effet and extending the test time, we tailor the driver gas mixture. Typially the driver is helium, if we add nitrogen the mixtures speed of sound is redued. We an reate a mixture in the driver that will have a gammaa exatly mathed to the gamma of the driver gas at the temperatures of region 2, 109

127 and in essene we an avoid refleting the refleted shok (Figure C.0.2). This only works to a point. The ontat surfae is not a step funtion but is a region of mixed gases so the end result is very small pressure osillation. Figure C.0.2: Pressure trae from pressure transduer loated at 2 m from the endwall on a test using a tailored driver mixture. At 0 ms the inident shok arrives followed losely by the refleted shok. In region 5 the pressure slowly inreases by some non-ideall faility effets. At 6 ms the refletion off the ontat surfae arrives but with no signifiant pressure jump. The pressure trae is reasonably level until the arrival of the expansion wave at 11 ms. C.4 57BC CONSTAC ANT U, V OPERATION Another faility-dependena slightly positive dp/dt (this effet an be seen in the first 6 ms of Figure effet that beomes important at long test-time onditions is C.5) aompanied by a dt/dt, whih after enoughh time effets the hemistry. This effet ours beause the boundary layer s visous dissipation slows the inident shok, and as a result region 2 (and onsequentia ally region 5) has a pressure profile that dereases 110

128 loser to the endwall. As the pressure profile equalizes near the endwall the pressure inreases. This hallenges the onstant UV assumptions that are very useful for modeling. This will artifiially make the measured ignition delay times our sooner than would be predited by a onstant UV assumption. See Figure C.0.3. Some attempts have been made to apture this non ideal effet in the model and thus prediting the roll off. But these are ompliated when there is signifiant heat release upon ignition. A more elegant solution involvess removing the rise in pressure by partially refleting the expansion wave prematurely using speially designed driver inserts to keep the pressure profile onstant. Another method of aounting for this pressure rise is to inlude it in the modeling as below in Figure C The model was modified by speifying a pressure that mathes the measured pressure. This is not ompletely aurate beause during ignition when a large amount of energy is released the pressure will rise, but it gives us some insight as to what onditions this effet beomes signifiant. Figure C.0.3: Figure 4.9 of ignition delay time plot vs. inversee of temperature showing the effet of modeling as onstant UV or with a 0.2 atm/ms dp/dt. The effet beomes apparent at test times above 10 ms.. 111

129 APPE 14BA ENDIX D 58BD.1 D DETAILED CALCUC ULATION OF PRESSURES INITIAL In pratie, a more omplex model than the isothermal model presented in Setion is used for setting the initial pressures for the seond generation aerosol shok tube. To help understand this more omplex model a very simplified representation of the expansion proess is shown in Figure D.0.1. Initially P A is larger than P B and a sliding piston, whihh separates the volumes V A and V B B, is held in plae. The piston is then allowed to move, (this is analogous to opening the ball valve) and the gas on the left is ompressed as the gas on the right expands until the pressures equalize to P 1. Figure D.0.1: Simplified shemati of the expansion proess showing (a) the initial state and (b) the final state. The volume V A represents the test setion andd mixing tank (V A = V TS +V MT ) and V B represents the dump tank from Figure 3.2. V T is the total displaed volume and is related to V (See Figure 3.2) by V T = (V TS /V MT 1)V.. This definition makes the equations simpler. 112

130 Expansion experiments were performed in the shok tube with a water aerosol. In these experiments, we kept the initial pressure in the aerosol mixing tank and test setion onstant (P A = 700 torr) and we varied the initial pressure in the dump tank (P B ). We then measured the resultant equilibrium pressure (P 1 1). The dataa are shown in Figure D.0.2. To model these results we first assumed the proess proeeded as in Figure D.0.1 isothermally. The isothermal model underpredits the pressure as seen in Figure D.0.2. Figure D.0.2: Measurements of final pressure (P 1 ) after expansionn of gas/aerosol mixture into dump tank. Initial pressure in the aerosol mixing tank and test setion (P A ) wass around 700 torr and the initial pressure in the dump tank (P B ) was varied. Also shown are various models used to predit the expansion behavior. The hybrid model that assumes isothermal expansionn and isentropi ompression seems to predit the final pressure quite well. We then tried an isentropi model, whih does a muh better job of prediting the final pressure. To better understand what happened inn our experiment, we made a temperature measurement in the ore of the dump tankk and the ore of the test setion near the endwall. We saw a slight inrease in temperature in the dump tank, whih would be expeted in an isentropi ompression, but no notieable hange in the temperature in the test setion during expansion. This is most likely due to the fat that the droplets regulatee the temperature by ating as ondensation sites. Beause of this 113

131 result, a hybrid proess would be more appropriate. More speifially, we used the isothermal assumption in the aerosol mixing tankk and test setion for the expansion, and an isentropi assumption in the dump tank forr the ompression. The final pressures predited using this hybrid model math most losely with the experimental data. Now that we know whih model is most aurate, we an use this hybrid model to find the initial pressures given a desired final pressure using the equations below. P B = ((V B -V T )/V B ) γ P 1 P A = P1(V A +V T )/V V A Equation. 3.1a Equation. 3.1b The pressures and volumes are defined in Figure 3.1 and Figure D.0.1. γ is the ratio of speifi heats for the bathgas. An example of this alulation is shown in Figure D.0.3. The initial pressures P A and P B are plotted on the left axis versus the desired P 1. If a 100 torr pre-shok pressure (P 1 ) is required, the aerosol mixing tank and test setion should be set to 185 torr and the dump tank should be set to 62 torr for a non-dimensional expansion of X = 2. Figure D.0.3: Plot indiating how to set initial pressures inn the mixing tank/test setion and the dump tank. For example, to ahieve a equilibrium pressure (P 1 ) of 100 torr and expansion fator of 2, the mixing plenum and testt setion should be set initially to 185 torr (P A ) and the dump tank should be set initially to 62 torr (P B ). 114