SOOT ABATEMENT USING FUEL ADDITIVES

Transcription

1 The Pennsylvania State University The Graduate School College of Engineering SOOT ABATEMENT USING FUEL ADDITIVES A Thesis in Mechanical Engineering by Juntao Wu 2004 Juntao Wu Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy December 2004

2 The thesis of Juntao Wu has been reviewed and approved* by the following: Thomas A. Litzinger Professor of Mechanical Engineering Thesis Advisor Chair of Committee Robert J. Santoro George L. Guillet Professor of Mechanical Engineering Richard A. Yetter Professor of Mechanical Engineering Harold H. Schobert Professor of Fuel Science Richard C. Benson Professor of Mechanical Engineering Head of Department of Mechanical and Nuclear Engineering *Signatures are on file in the Graduate School

3 iii ABSTRACT A study of the mechanism of soot abatement by oxygen-containing additives was conducted by both experimental and modeling methods. The overall technical objective is to develop a fundamental understanding of the complex roles of oxygen-containing additives in the processes which lead to particulate matter emissions. Two classes of compounds were investigated - oxygenates and nitro-alkanes. Experiments were performed on a premixed ethylene/air flame with two oxygenated additives - ethanol and dimethyl ether (DME). The experiments were conducted at two equivalence ratios, φ = 2.34 and φ = 2.64, and two levels of ethanol or DME, 5% and 10% oxygen in mass of the fuel, were added to the fuel at each equivalence ratio. The experimental results show that both ethanol and DME reduced aromatic species and soot, and they were more effective at φ = 2.34 than at φ = The comparison between ethanol and DME shows that, at Ф = 2.34, DME is more effective than ethanol on PAH and soot reduction; however, at φ = 2.64, there is no detectable difference between these two additives on the aromatic species and soot suppression. A chemical model, Howard-DME-Ethanol (HDE) mechanism, was used to investigate the chemical processes leading to the abatement of aromatic species and soot by ethanol and DME. The analysis shows that these reduction effects result from the removal of carbon from the pathway to aromatic species formation. DME is more

4 iv effective than ethanol in reducing soot precursors, because more carbon in DME is removed from the participation of aromatic formation than that in ethanol. This difference is due to the molecular structure difference between ethanol and DME. Screening studies of the nitro-alkanes showed them to be effective in reducing soot. The primary mechanism of soot reduction was hypothesized to be linked to NO 2, so a modeling study of NO 2 addition to premixed flames was undertaken. A chemical model, Howard-NO 2 (HN) mechanism, was used to investigate the chemical processes leading to the effect of NO 2 on aromatic species and soot reduction. The mechanism analysis shows that the addition of NO 2 increases the level of OH radicals in the flame, through reactions of NO 2 +H NO+OH, HO 2 +NO NO 2 +OH. Then the increased OH decreases the level of H 2 by reactions H 2 +OH H 2 O+H. The lower level of H 2 increases the reaction rate of H 2 CCCH+H C 3 H 2 +H 2, and results a lower level of H 2 CCCH. Since all C 6 H 6 comes from H 2 CCCH through reaction 2H 2 CCCH C 6 H 6, low level H 2 CCCH leads to a reduction of C 6 H 6. As a key element during the aromatic species growth, low level of C 6 H 6 leads to suppression of overall aromatic species formation and growth.

5 v TABLE OF CONTENTS ABSTRACT...iii TABLE OF CONTENTS... v LIST OF FIGURES...viii LIST OF TABLES...xiii NOMENCLATURE... xiv ACKNOWLEDGEMENTS... xvi Chapter 1 INTRODUCTION Motivation Objectives Soot Formation Soot Particle Precursor and Particle Nucleation Particle Coagulation Soot Growth Soot Oxidation General Soot Physical Properties Soot Reduction by Additives Inert Diluents Gaseous Additives Metallic Additives Oxygenated Additives Chapter 2 PRINCIPLES OF DIAGNOSTIC METHODS Laser-induced Incandescence Theoretical Analysis Experimental Considerations Laser Excitation: Intensity Profile, Energy, and Wavelength Spectral Detection Region Detection Gate Width and Timing Calibration Calibration via Laser Extinction Calibration via Cavity Ringdown... 50

6 vi Calibration via Gravimetric Techniques Calibration via 3-Color Pyrometry General Concerns in Calibrating LII Laser Extinction Laser-induced Fluorescence Theory of Laser-induced Fluorescence LIF Calibration Temperature Measurement Chapter 3 EXPERIMENTAL APPROACH Materials Experimental Approach Chapter 4 EXPERIMENTAL RESULTS AND MODELING ANALYSIS FOR ETHANOL AND DME Experimental Results Temperature Profiles Spectra Scanning and PAH Concentration Profiles Soot Volume Fraction Profiles Effect of Additives on PAH and Soot Volume Fraction Profiles Modeling Analysis DME Consumption Ethanol Consumption Effect of Molecular Structure of Additives Effect of Temperature Chapter 5 MODELING STUDY OF NO 2 ON SOOT REDUCTION Modeling Conditions Chemical Mechanisms Validation of Combined Mechanisms Reaction Pathways Fuel Destruction to Benzene Ethylene Consumption Benzene Formation PAH Growth Naphthalene (C 10 H 8 ) Formation Phenanthrene (A3) Formation

7 vii Pyrene Formation Mechanisms of Soot Reduction Overview Chemical Effect of NO NO 2 Reactions Effect of NO 2 on Benzene Formation Effect of NO 2 on PAH Growth Effect of Temperature Temperature Effect on Benzene Formation Temperature Effect on C 8 H 6 Formation Chapter 6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE WORK Conclusions Suggestions for Future Work REFERENCES APPENDIX Reactions for HN mechanism...229

8 viii LIST OF FIGURES Figure 1-1 Jet-Propulsion Fuel Figure 1-2 Laboratory flames and reactors with associated chemical and physical processes Figure 1-3 TEM photograph of typical soot aggregates emitted form long residence time turbulent acetylene/air diffusion flames where emitted soot properties were independent of residence time Figure 1-4 Total surface area of soot particles per cm 3 = π <d 2 >N, as a function of age for flames of 3 stoichiometries:, C/O=0.79;,0.76;, Figure 1-5 Upper portion: volume fraction of soot as a function of time for flames with six different stoichiometries, as shown. The approximately 15 data points for each curve are omitted for clearity. Lower portion: specific surface growth rates as a function of time Figure 1-6 Effect of the relative reactivity of soot and ΣPAH. C/O=0.79. ( ) 10mm,(Δ) 17mm, ( ) 25mm Figure 1-7 Collision efficiencies at different distances from burner. γ acetylene-soot ( 10 4 ) : ( ) C/O = 0.70, ( ) C/O = 0.79; γ PAH-soot : ( )C/O = 0.70,( ) C/O = Figure 1-8 Sketch of the structure of soot emitted from a diesel engine Figure 1-9 Influence of various gaseous additives on the critical C/O-ratio for the onset of sooting in an ethylene/air flame Figure 1-10 Influence of gaseous additives on the downstream ( t 30 ms ) soot yield in an ethylene/air flame. C/O ~ 0.76, fv~ 1.5 x Figure 1-11 Smoke emissions from diesel engines with oxygenated additives, from Miyamoto et al Figure 1-12 Acetylene concentration evolution of fuels with about same centane number. In the figure, intervals of confidence of 80% are indicated. Test condition: engine speed = 1250 rpm, 4.5 bar of i.m.e.p., start of combustion = TDC Figure 2-1 Optical setup for 2-D LII measurements. LDF laminar diffusion flame, BB beam block, DM dichroic mirror, ½-λ - half-wave plate, GL Glan-Laser prism, A aperture, P prism, CL cylindrical lens, SL spherical lens, and ICCD intensified CCD Figure 2-2 Fluence dependence of LII measured in steady laminar diffusion flames. Data were collected at H = 20 mm in the ethylenec/air flame for detection gate durations of 19 ns (+) and 85 ns ( ) both gates opening coincident with the arrival of the 5 ns laser pulse. Data are also shown for the methane/air flame at H = 50 mm with 85 ns gate ( ). Raw signals for each condition have been normalized to a value of 1.0 at a fluence of 0.6 J/cm 2. The solid line shown is the least-squares power-law fit of the methane data for fluences greater than 0.03 J/cm 2 ; the fit follows the expression signal fluence reference [89]

9 ix Figure 2-3 Temporal profile of a LII signal obtained in the ethene-air laminar diffusion flame at heights of 10 and 30 mm above the fuel tube exit and at the radial locations corresponding to peak soot volume fraction for these heights Figure 2-4 Extinction by a collection of particles Figure 2-5 Schematic diagram of two-level model of induced fluorescence Figure 3-1 Chemical structure of ethanol and DME Figure 3-2 Schematic diagram of experimental setup Figure 3-3 Schematic diagram of fuel, additive, and air supply systems Figure 3-4 The size of PAH and their spectral properties Figure 3-5 Diagram of the Thermocouple Arrangement Figure 4-1 Temperature profiles of premixed ethylene/air flame, with and without ethanol and DME addition Figure 4-2 Flame temperature profiles from Intal and Senkan Figure 4-3 Pictures of the ethylene/air flames with and without additives. (Digital Camera Setting: F=2.8, Time=1/800 s) Figure 4-4 Fluorescence spectra. (a) φ = 2.34; (b) φ = Figure 4-5 Fluorescence spectra of laminar premixed ethylene/air flat flame at different heights above the burner surface Figure 4-6 Normalized fluorescence intensity profiles of small PAH Figure 4-7 Normalized fluorescence intensity profiles of large PAH Figure 4-8 Comparison of soot volume fraction measurement results Figure 4-9 The effect of ethanol on small PAH profiles Figure 4-10 The effect of ethanol on large PAH profiles Figure 4-11 The effect of ethanol on soot profiles Figure 4-12 Comparison of ethanol effects on soot volume fraction profiles at Ф= Figure 4-13 The effect of DME on small PAH profiles Figure 4-14 The effect of DME on large PAH profiles Figure 4-15 The effect of DME on soot profiles Figure 4-16 Comparison of the effects of ethanol and DME on small PAH profiles Figure 4-17 Comparison of the effects of ethanol and DME on large PAH profiles Figure 4-18 Comparison of the effects of ethanol and DME on soot profiles Figure 4-19 Comparison of the effects of ethanol and DME on predicted small PAH profiles Figure 4-20 Comparison of the effects of ethanol and DME on predicted large PAH profiles Figure 4-21 Comparison of the effects of ethanol and DME on predicted soot profiles108 Figure 4-22 Reaction flux diagram of DME consumption Figure 4-23 Reaction flux diagram of CH 3 consumption with 5% oxygen in the fuel by the addition of DME at Ф=

10 x Figure 4-24 Reaction flux diagram of CH 3 consumption with 5% oxygen in the fuel by the addition of DME at Ф= Figure 4-25 Reaction flux diagram of ethanol consumption with 5% oxygen in the fuel by the addition of ethanol at Ф= Figure 4-26 Reaction flux diagram of CH 3 consumption with 5% oxygen in the fuel by the addition of ethanol at Ф= Figure 4-27 Reaction flux diagram of ethanol consumption with 5% oxygen in the fuel by the addition of ethanol at Ф= Figure 4-28 Reaction flux diagram of CH 3 consumption with 5% oxygen in the fuel by the addition of ethanol at Ф= Figure 4-29 Mole concentration profiles of pyrene calculated using original and adjusted temperature profiles Figure 5-1 Comparison between Howard and HN mechanisms at Φ = Figure 5-2 Comparison between Howard and HN mechanisms at Φ = Figure 5-3 Small PAH profiles Figure 5-4 Large PAH profiles Figure 5-5 Soot volume fraction with and without NO 2 at two equivalence ratios Figure 5-6 Reaction flux diagram of C 2 H 4 consumption at Ф= Figure 5-7 Reaction flux diagram of C 2 H 4 consumption at Ф= Figure 5-8 Reaction flux diagram of benzene production and destruction at Ф= Figure 5-9 Net production rate of benzene at Ф = Figure 5-10 Net production rate of benzene at Ф = Figure 5-11 Reaction flux diagram of C 10 H 8 production and destruction at Ф= Figure 5-12 Reaction flux diagram of C 10 H 8 production and destruction at Ф= Figure 5-13 Reaction flux diagram of A3 production and destruction at Ф= Figure 5-14 Reaction flux diagram of A3 production and destruction at Ф= Figure 5-15 Reaction flux diagram of pyrene (C 16 H 10 ) production and destruction at Ф= Figure 5-16 Reaction flux diagram of pyrene (C 16 H 10 ) production and destruction at Ф= Figure 5-17 Carbon flux among PAHs at Ф= Figure 5-18 Carbon flux among PAHs at Ф= Figure 5-19 Flux diagram of nitrogen atoms introduced by NO 2 at Ф= Figure 5-20 Flux diagram of NO 2 consumption at Ф= Figure 5-21 Flux diagram of NO 2 consumption at Ф= Figure 5-22 Mole concentration profiles of OH at Ф= Figure 5-23 Mole concentration profiles of OH at Ф= Figure 5-24 Mole concentration profiles of benzene Figure 5-25 Net production rate of C 6 H 6 at Ф= Figure 5-26 Net production rate of H 2 CCCH at Ф= Figure 5-27 Net production of H 2 CCCH at Ф= Figure 5-28 Total production and consumption of H 2 CCCH at Ф=

11 Figure 5-29 Mole concentration profiles of H 2 CCCH at Ф= Figure 5-30 Mole concentration profiles of H at Ф= Figure 5-31 Mole concentration profiles of C 3 H 2 at Ф= Figure 5-32 Mole concentration profiles of H 2 at Ф= Figure 5-33 Net production rate of H 2 at Ф= Figure 5-34 Net production of H 2 at Ф= Figure 5-35 Total production and consumption of H 2 at Ф= Figure 5-36 Mole concentration profiles of H 2 CCCH at Ф= Figure 5-37 Mole concentration profiles of H at Ф= Figure 5-38 Mole concentration profiles of C 3 H 2 at Ф= Figure 5-39 Mole concentration profiles of H 2 at Ф= Figure 5-40 Mole concentration profiles of C 8 H Figure 5-41 Net production rate of C 10 H 8 at Ф= Figure 5-42 Net production of C 10 H 8 at Ф= Figure 5-43 Total production and consumption of C 10 H 8 at Ф= Figure 5-44 Net production rate of A3 at Ф= Figure 5-45 Net production of A3 at Ф= Figure 5-46 Total production and consumption of A3 at Ф= Figure 5-47 Net production rate of pyrene at Ф= Figure 5-48 Net production of pyrene at Ф= Figure 5-49 Total production and consumption of pyrene at Ф= Figure 5-50 Effect of temperature on benzene with NO Figure 5-51 Effect of temperature on C 10 H 8 with NO Figure 5-52 Effect of temperature on A3 with NO Figure 5-53 Effect of temperature on pyrene with NO Figure 5-54 Effect of temperature on H 2 CCCH with NO Figure 5-55 Effect of temperature on net production rate of benzene at Ф=2.34 with NO Figure 5-56 Effect of temperature on net production of benzene at Ф=2.34 with NO Figure 5-57 Effect of temperature on total production and consumption of benzene at Ф=2.34 with NO Figure 5-58 Reaction constant curve as a function of temperature Figure 5-59 Effect of temperature on H 2 at Ф=2.34 with NO Figure 5-60 Effect of temperature on C 6 H 5 at Ф=2.34 with NO Figure 5-61 Effect of temperature on H at Ф=2.34 with NO Figure 5-62 Effect of temperature on H at Ф=2.64 with NO Figure 5-63 Effect of temperature on benzene at Ф=2.64 with NO Figure 5-64 Effect of temperature on C 8 H 6 with NO Figure 5-65 Effect of temperature on net production rate of C 8 H 6 at Ф = 2.34 with NO Figure 5-66 Effect of temperature on C 6 H 5 consumption at Ф = 2.34 with NO Figure 5-67 Reaction constant curve as a function of temperature Figure 5-68 Effect of temperature on net production rate of C 6 H 5 (L) from reaction C 6 H 5 C 6 H 5 (L) xi

12 xii Figure 5-69 Effect of temperature on reaction rate of HCCHCCH from reaction C 6 H 5 HCCHCCH+C 2 H Figure 6-1 Fluorescence profiles with and without NO

13 xiii LIST OF TABLES Table 2-1 Refractive indices from Chang and Charalampopoulos Table 3-1 Test Conditions Table 4-1 The reduction ratios by the addition of ethanol at Ф=2.34 and Table 4-2 The reduction ratios by the addition of DME at Ф=2.34 and Table 4-3 The reduction ratios by the addition of ethanol and DME at Ф=2.34 and Table 5-1 Modeling conditions Table 5-2 Major reactions involved in NO 2 consumption and production at Ф= Table 5-3 Adiabatic flame temperature prediction result from CEA Table 5-4 Arrhenius rate coefficients of reaction C 6 H 6 +H C 6 H 5 +H Table 5-5 Total C 6 H 6 production and consumption at Ф=2.34 and Table 5-6 C 8 H 6 production from C 6 H Table 5-7 C 6 H 5 consumption Table 5-8 Arrhenius rate coefficients of reaction C 6 H 5 C 6 H 5 (L)

14 xiv NOMENCLATURE A1 benzene A1- phenyl A1C 2 H A2 phenylacetylene naphthalene A2-1 naphthyl-1 A2-2 naphthyl-2 A3 A3L A4 AC 3 H 4 A1C 2 H*2 phenanthrene anthracene pyrene allene phenylacetylene A3-4 phenanthryl-4 A3*4 phenanthryl-4 C 3 H 4 P C 10 H 7 *1 C 10 H 7 *2 C 10 H 7 O-1 C 10 H 7 O-2 INDENE* Nu propyne naphthyl-1 naphthyl-2 naphthoxy-1 naphthoxy-2 indenyl Nusselt Number

15 xv Re P2 Pr PYRENE T Reynolds Number biphenyl Prandtl Number pyrene Temperature in Kelvin [K] Greek Letters ε g Emissivity Gas µ Dynamic Viscosity Subscripts d tc Diameter Thermocouple Junction

16 xvi ACKNOWLEDGEMENTS Many people have contributed to the accomplishment of this work. First and foremost, I would like to thank my advisor Professor Thomas Litzinger for his constant support, for many inspiring discussions, for being an inexhaustible source of new ideas, and for his patience throughout my graduate study. His encouragement and help in academic as well as non-academic matters, made the whole experience of my graduate studies at the PennState University highly enjoyable and educational. It was an honor to have advices from someone like Professor Robert Santoro who has a great reputation in combustion society. I always appreciate his comment and advice with deep knowledge in combustion science. I also want to thank Professor Richard Yetter, and Professor Harold Schobert for their time and effort in directing my thesis work. I am grateful to Professor Dan Haworth, for allowing me to use his computing equipment. He always helped me whenever I had troubles with computational work. I also want to thank Professor Stephen R. Turns and Professor Domenic Santavicca, for their help and fruitful discussions. I also want to thank people who worked with me and helped me: Dr. Hyungsik Lee, Dr. Seong-Young Lee, Dr. Ki Hoon Song, Dr. Milt Linevsky, and Dr. Jianquan Li. I wish them the best of luck in all of their endeavors. Finally, I would like to thank my parents and wife, who are always the most important part of my life. Thanks for their love, understanding, and support.

17 1 Chapter 1 INTRODUCTION 1.1 Motivation The ramification of soot formation in practical combustion systems has long been identified to have a harmful effect on the environment, public health, and hardware performance. In epidemiological studies 1,2, air pollution was positively associated with death from lung cancer and cardiopulmonary disease. Fine particles defined by a diameter equal to or below 2.5 µm are thought to pose a particularly great risk to health because they are more likely to be toxic than larger particles and can be breathed more deeply into the lungs. 3 The observation that soot can cause cancer of the skin in man was first made by Pott in The polynuclear aromatic hydrocarbons (PAH) absorbed on soot are believed to be responsible for this effect, 5 as many PAH are known to be carcinogenic. 6 As for hardware performance consideration, turbine combustor performance is usually limited by the temperatures which the construction materials can withstand. 4 Increased emissivity and higher radiative heat transfer caused by soot formation in the combustor can cause overheating and damage. Excessive quantities of soot particles can erode turbine blades and cause carbonaceous deposits leading to fuelspray distortion. These problems would become even more serious with any trend toward more aromatic fuels. 7 Militarily, a fuel-rich primary combustion zone is desirable as it

18 2 improves high altitude relight capability. However, a sooty exhaust trailing jet aircraft is equally undesirable since it enhances detection by an adversary. 8 It is estimated that US military aircraft emit about 600,000 kg of particulate matter (PM) into the atmosphere each year. Without control, this PM can cause serious environment pollution and results in a direct health threaten to human body. These concerns have motivated many studies focused on understanding and control of soot formation. In the past two decades, much work has been done on diesel engines, and soot emission from diesel engine has been steadily reduced driven by the stricter environment standards. One of the major factors in this reduction has been changing the fuel composition including reduction of sulfur levels and aromatic species. However, these approaches are not likely to be viable for gas turbines. It is well established that the sulphur content has a direct impact on the emission of particles. 9 Currently the JP-8 sulphur content is limited to a max of 0.3 wt%. Actually, a recent study has shown that the typical average sulphur content in aviation kerosene is wt%, see Figure 1-1. A further reduction of the sulphur content, however, would lead to diminishing returns, i.e. in Diesel a reduction from 0.2 to 0.05 wt% of sulphur yields up to 22% reduction in particulate emissions, whereas a further reduction from 0.05 to 0.02 wt% only gives a further 4% reduction. Furthermore, this is confirmed by the use of sulphur free synthetic Fischer-Tropsch fuels, produced in a gas-to-liquid process, which also leads to the emission of particles.

19 3 Sulfur C ontent in JP-8, 1997 B uys ppm Million gallons ppm Sulfur, M ass % Figure 1-1 Jet-Propulsion Fuel 8. 10

20 4 Aromatic hydrocarbons have the greatest soot-forming potential of fuel hydrocarbons, so reducing the aromatic hydrocarbon content would significantly reduce the emissions of particles. Specifying a 25% per volume maximum limit, with no minimum level specified normally controls the aromatic hydrocarbon content in aviation fuels. One of the drawbacks when significantly reducing the aromatic hydrocarbon content of the fuel is the interaction between the fuel and elastomeric seal materials. These materials, which are used widely in the fuel system, swell to varying degrees, which depend strongly on the aromatic hydrocarbon content of the fuels. When the aromatic hydrocarbon content is reduced, the shrinking of the seals can lead to fuel leakages. Therefore, for compatibility, a minimum aromatic content of 10% by volume is required. In addition, aromatic species are required to maintain acceptable freezing point for the fuel. Using oxygenated additives to reduce soot has currently being studied intensively. While there is much ambiguity regarding the dominant factor in the control of soot emissions for oxygenated fuels, many studies have demonstrated that they reduce particulate emissions substantially (Liotta 1993; Ullman 1994; Spreen 1995; Tsurutani 1995; Beatrice 1996; Neeft 1996; Bertoli 1997; Edgar 1997; McCormick 1997; Grabowski 1998; Hess 1998; Miyamoto 1998; Choi 1999: Stoner 1999; Litzinger 2000; Babi 2000). At the same time, for existing engines, hardware retrofits are normally prohibitively expensive. Therefore, the initial approach to reduce emissions from gas turbines that was selected for study in this program was the introduction of oxygenated compounds into the fuel.

21 5 1.2 Objectives The work presented was conducted as part of larger effort of a consortium of DoD laboratories, universities, and industry seeking fuel additives to reduce particulate emissions from military gas turbine engines. 11 Compared with fuel or engine hardware changes, the use of fuel additives is a pervasive and cost effective approach that has the potential of reducing particulate emissions in all engines in the fleet. The objectives of the overall research program are to develop fundamental understanding of the complex interactions of additives with the processes that lead to PM emissions from military gas turbine engines and to use that fundamental understanding to select and investigate the most promising additives for reducing PM emissions. The overall research program involves testing in premixed and diffusion flames, a well-stirred reactor, a shock tube, an atmospheric spray burner, and a high pressure, turbulent combustion reactor. A complete set of laboratory devices and the associated hierarchy of chemical and physical processes are illustrated in Figure 1-2. Understanding the effects of additives across this array of devices will permit a more complete understanding of the effects of additives in gas turbine combustors. Our specific research work is focused on premixed flame. Premixed flame provides a relatively stable environment and is ideal for experimental investigation. Ethylene was selected as the fuel due to its use in many previous studies of soot formation in a variety of combustion devices. Another consideration in the selection of ethylene was the availability of chemical kinetic mechanisms for simulation. Thus experiments were performed on a premixed ethylene/air flame. Ethanol and dimethyl

22 6 ether were chosen as oxygenated fuel additives because they are well-known for effectively reducing soot emission and have relatively simple chemical structure. In order to identify the role of oxygenates in soot reduction, modeling analysis was conducted. Detailed kinetic modeling study provides deeper insights into the chemical processes responsible for soot formation and oxidation that generally can not be achieved through the experimental studies, and therefore allows identification of the role of oxygenates in fuel during combustion. Increasing Complexity Kinetics +Diffusion +Turbulent mixing & Sprays + Bulk Mixing Well-stirred Reactor Shock tube Premixed Flame Diffusion Flame (opposed-jet type) Turbulent Spray Flame High Pressure Combustion Reactor Liquid fuel Gaseous fuel Figure 1-2 Laboratory flames and reactors with associated chemical and physical processes. 12 Another task of present study is to scrutinize the detailed soot reduction mechanism of NO 2 through modeling investigation. Recent testing at AFRL (Air Force

23 7 Flight Research Laboratory) has identified a commercial additive that can significantly reduce soot emissions from a gas turbine engine under certain conditions. This is a commercial additive and its constituents are known. The major components of this additive that are expected to reduce soot are nitro-methane, nitro-ethane, and nitropropane. However, the particulate suppression mechanism of this additive has not been established. Because the initial step of nitroalkane decomposition was the bond scission 13, 14, 15 of C-N forming NO 2 and alkane, NO 2 is thought to play the key role in understanding the effect of nitro-alkanes on soot reduction. Though the produced alkane may also affect the production of soot, present chemical mechanism has a limitation to simulate the effect of C 2 H + 5 and C 3 H + 7, therefore, our work was focused on the effect of NO 2, and NO 2 was chosen as an additive in premixed Ethylene/air flame to study the soot reduction mechanism of nitro-alkanes. The experimental work was conducted by a former researcher, 16 and the task of present study is to reveal the detailed soot reduction mechanism of NO 2 through modeling investigation. A potential downside to the use of nitro-compounds to reduce soot is that they could lead to increased levels of NOx emissions from engines. The extent to which the nitro-alkanes would increase NOx depends on the level at which they must be added to the fuel and also the fraction of the NO that survives the combustion reactions that convert it to nitrogen under fuel rich conditions. As background information for this research program, soot formation processes, general soot properties, and soot reduction mechanisms by various additives, especially the oxygenated additives, are discussed in following sections.

24 8 1.3 Soot Formation The progress of soot formation in flames may be divided into four major processes: particle precursor nucleation, particle surface growth, particle coagulation, and oxidation Soot Particle Precursor and Particle Nucleation Several principal proposals are made to reveal the nature of soot particle inception. Polyacetylenes, 17 ionic species, 18 polyynes, 19,20 and polycyclic aromatic hydrocarbons 21 are all potential gaseous precursors of soot. Homann and Wagner proposed that polyacetylenes were important soot precursors, based on their mass spectral experiment. 17 Polyacetylene model is easy to account for the formation of large fraction of polyacetylenes observed. When large molecules form, the average size of the radicals also continues to increase, and large radicals react with themselves and with bigger polyacetylenes, forming even larger molecules. However, Cullis et al. doubted this model because they thought that the polyacetylenes did not grow sufficiently rapidly to account for the almost instantaneous formation of soot. 22 Based on Cullis s argument, Calcote claimed further that the reactions of neutral species were not fast enough and hence suggested an ionic mechanism. He assumed

25 9 chemiions to be the precursor on which free radicals, polyacetylenes, and PAH repeatedly add through fast ion-molecule reactions. 18 Calcote claimed that H 3 O + was dominant in + near stoichiometric and lean flames, and C 3 H 3 was the dominant ion in rich flames. He stated that the primary ions were produced by following chemiionizations: CH + O CHO + + e - (1.1) CHO + + C 2 H 2 O C 2 H 3 O + + CO (1.2) C 2 H 3 O + + C 2 H 2 C 3 H CH 2 O (1.3) CHO + + H 2 O H 3 O + + CO (1.4) Reactions (1.1), (1.2), and (1.3) are representatives of a number of mechanisms going from CHO + to C 3 H + 3, and reaction (1.4) is the source of H 3 O +. Calcote stated that as ionmolecule reactions were so quick that CHO + could not be observed in large concentration. Instead of ion, however, Krestinin proposed that polyynes could be the gaseous precursors of soot, based on the high reactivity of these species in polymerization reactions. 19, 20 Krestinin argued that the multi-stage increases in the number of aromatic rings in the PAH were rather slow, while the polyynes (C 2n H 2 ) grew in a simple and fast way, typically in C 2n H 2 + C 2 H = C 2n+2 H 2 + H (1.5) consequently, the concentrations of the polyynes C 4 H 2, C 6 H 2, C 8 H 2 were high enough to be detected as major hydrocarbon intermediates in both pyrolysis and combustion processes.

26 10 In spite of the fast reaction speed of above mechanisms, however, the majority opinion at present is PAH hypothesis, supported by numerous experimental and modeling studies. 23,24 In PAH model, the formation of the first aromatic ring from small aliphatics is emphasized because this process is believed to be the rate-limiting step in reactions leading to larger aromatics. Frenklach et al. 19 suggested reaction n-c 4 H 3 + C 2 H 2 phenyl (1.6) played a key role in the formation of the first aromatic ring, based on detailed kinetic simulations of shock-tube acetylene pyrolysis; Bittner and Howard 25 suggested reaction n-c 4 H 5 + C 2 H 2 benzene + H (1.7) played a role at lower temperature, which was also identified by the Frenklach et al. s simulation. However, Miller and Melius 26 dismissed reactions (1.6) and (1.7). They argued that n-c 4 H 3 and n-c 4 H 5 could not be presented in sufficiently high concentrations because they transformed rapidly to their corresponding resonantly stabilized isomers, iso-c 4 H 3 and iso-c 4 H 5. Instead, they proposed reaction C 3 H 3 + C 3 H 3 benzene or phenyl + H (1.8) since propargyl is an exceptionally stable hydrocarbon radical, and its implication in the 27, 28 formation of aromatics and soot has long been assumed. Frenklach argued that the stability of n-c 4 H 3 and n-c 4 H 5 was actually higher than that estimated by Miller and Melius. What s more, Frenklach et al. provided another possibility for the initial ring formation, reaction

27 11 C 3 H 3 + C 2 H 2 c-c 5 H 5 (1.9) Propargyl reacts with acetylene to form a cyclopentadienyl radical. This pathway combines the benefits of the two reactants types discussed above: highly stable radical, propargyl, and the most abundant building block, acetylene. Once formed, cyclopentadienyl reacts rapidly to form benzene. 29,30,31 Based on the comparison of rate coefficients between reactions (1.9) and (1.8), Frenklach et al. claimed that reaction (1.9) was expected to be faster than reaction (1.8) by a factor of 2 to This implies that reaction (1.8) is not only fast enough to make a difference, but also probably plays a dominant role in the formation of first aromatic ring. After the first aromatic ring is formed, it will develop into bigger rings. Frenklach et al. proposed an HACA mechanism to explain for this growth. HACA was introduced in Frenklach and Wang s paper, 32 as an acronym for H-Abstration-C 2 H 2 -Addition, which implied a repetitive reaction sequence of two principal steps: (i) abstraction of a hydrogen atom from the reacting hydrocarbon by a gaseous hydrogen atom, A i + H A i - + H 2 (1.10) followed by (ii) addition of a gaseous acetylene molecule to the radical site formed, A i - + C 2 H 2 products (1.11) Here, A i is an aromatic molecule with i peri-condensed (peri: enclosing) rings, and A i - is its radical.

28 12 The essence of the two-step feature is that the first step activates a molecule to further growth by converting it to a radical, which is reversible. The reverse steps can be the reverse direction of the H abstraction itself, A i - + H 2 A i + H (1.12) or other reactions, such as the combination with a gaseous H, A - i + H A i (1.13) Frenklach et al. claimed that the contribution of reaction (1.13) increased with pressure and molecular size, and the reversibility of the acetylene addition step, reaction (1.11), was what determined whether this step would contribute to molecular growth Particle Coagulation After soot particles are formed, they collide with each other forming larger particles. Electron Transmission Microscope photos (Figure 1-3) showed that the shape of soot particles was initially spherical and changed into chain-like structure later. 21 Regarding this changing of shape, Prado et al. suggested that the particles were composed of viscous matter (liquid droplets) that coalesced completely at small sites but did not have sufficient time for fusion as the particle size increased; 33 Howard et al. believed that the nearly spherical shape of the primary particles was the product of simultaneously occurring coagulation and surface growth, and the transition to the chainlike aggregates was caused by cessation of surface growth. 34 Mitchell et al., using numerical simulations, showed another different picture. 35 Their simulations

29 13 demonstrated two factors affecting the level of particle sphericity. The first is a sufficiently fast surface growth rate. They claimed that for the geometry to become spherical, the rate of surface growth must be capable of burying colliding particles stuck to the surface of larger particles. When the colliding particles become large, the surface growth may not be able to bury them quickly enough. Thus, the second factor is the size of the colliding particles. Smaller particles are more easily covered. Figure 1-3 TEM photograph of typical soot aggregates emitted form long residence time turbulent acetylene/air diffusion flames where emitted soot properties were independent of residence time. 21

30 Soot Growth Acetylene (C 2 H 2 ) and PAH are two main potential agents responsible for soot surface growth. Soot growth by C 2 H 2 was promoted by Harris and Weiner after studying several premixed C 2 H 2 -air flat flames. 36 They concluded that only C 2 H 2 satisfied the requirements for a soot growth reactant and proposed a simple model in which soot mass growth rate was proportional to soot surface area and C 2 H 2 concentration. PAH were not measured because they were believed to have insufficient concentrations, and could not be counted as possible soot growth reactants. During their investigation of premixed ethylene/air flames at 6 different equivalence ratios ranged from φ = 2.1 to φ = 2.4, Harris and Weiner found that even though the soot volume fraction increased dramatically with equivalence ratio, the specific surface growth rates were almost the same. From scattering and extinction experiments, Harris and Weiner obtained the number density of soot particles N and their mean diameter <d> = Σ d i n i, where n i is the mole fraction of particles with diameter d i. They calculated the total surface area of soot particles per cm 3, using the definition of S = π<d 2 >N, and the result is showed in Figure 1-4. Also, their experimental results of soot volume fraction at different equivalence ratios are showed in the upper part of Figure 1-5. The authors attributed all of the increase in mass in those flames to surface growth since agglomeration did not affect the soot volume fraction. They differentiated these curves graphically and obtained surface growth rates. Although all the curves appear to have

31 15 similar slopes when plotted on semi-log coordinate, the derivatives are actually quite different and increase substantially when the equivalence ratio increases. Then, the authors divided the surface growth rates by the corresponding total soot surface areas showed in Figure 1-4, and converted volume to mass by multiplying a density of 1.8 g/cm 3. The resulting specific surface growth rates (g/cm 2 -s), are plotted as a function of time in the lower half of Figure 1-5. Soot Surface Area (cm 2 /cm 3 ) Time (ms) Figure 1-4 Total surface area of soot particles per cm 3 = π <d 2 >N, as a function of age for flames of 3 stoichiometries:, C/O=0.79;,0.76;, Figure 1-5 Upper portion: volume fraction of soot as a function of time for flames with six different stoichiometries, as shown. The approximately 15 data points for each curve are omitted for clearity. Lower portion: specific surface growth rates as a function of time. 36

32 16 The interesting feature of this result is that the specific surface rates for the four richest flames are within ±20 percent before 30 ms, beyond which the authors explained that the growth rates were too low to have enough accuracy. Also, they claimed that the greatest deviations in the specific surface growth rate were due to the experimental difficulty because the diameters of the soot particles were too small. Therefore, the most important result showed by Harris and Weiner is that the specific surface growth rate is only weakly dependent on stoichiometry between φ = 2.1 and φ = 2.4, compared with the total growth rate. The authors concluded that the much higher total growth rate of soot in richer flames was almost entirely due to the increased surface area available, and the concentration of growth species was similar in all of the flames, which is confirmed by Xu and Faeth s experimental data obtained at similar condition. 37 Harris and Weiner extended their conclusion and claimed that there was no depletion of growth species by surface growth. They ratiocinated that when the concentration of growth species was nearly independent of equivalence ratio, and the growth species concentrations were high enough to supply all the mass increase in the richest flame, if there is a depletion of growth species in the richest flame, then there must be an excess of growth species in the leaner flame, where the increase in soot mass was much less. Therefore, there could be no depletion of growth species in the richest flames. Furthermore, since the specific growth rate in the leaner flame was the same as that in the richest flame, there was no depletion of growth species in the leaner flame, either. Thus, Harris and Weiner ruled out the depletion of growth species.

33 17 Harris and Weiner also argued that althought the specific rates were similar in flames, the amounts of soot produced are quite different, so the processes determining the ultimate soot loading must have occured prior to the growth stage; that is, richer flames produce more soot because they have a higher nucleation rate and therefore larger surface area from the beginning of the growth stage. As a result, they ruled out PAH as the main soot growth species by arguing that: first, the PAH concentration changed sharply with stoichiometry, violating their conclusion that the concentration of growth species did not depend strongly on the stoichiometry of the flames; second, PAH concentrations were about 100 times higher in benzene flames than in flames of aliphatic fuels, 38 but the soot growth rates in both flames were similar; 39 third, there was always much less PAH material than soot in the flame, 40 so there could not be enough of it to account for the nearly order-of-magnitude increase in soot mass with age. Finally, Harris and Weiner considered acetylene as the dominant growth species because its concentration is high enough to account for the mass increase provided by surface growth. They also claimed that all of the soot came from, directly or indirectly, the acetylene in burned gases. Despite Harris and Weiner s agreements, Howard et al. did not agree with the above conclusions. Howard et al. believed that PAH played an important role in ethylene combustion. First, Howard et al. questioned the data from which Harris and Weiner first formulated their acetylene model. Howard et al. 41 repeated the particular flames conducted by Harris and Weiner 36 and found that previous soot concentration profiles for

34 18 the C/O = 0.79 flame was three times higher than their measurement results, which was in excellent agreement with interpolated values from optical measurement of Feitelberg 42 in similar ethylene flames at C/O = 0.77 and Meanwhile, in order to investigate the role of PAH in soot growth, Howard et al. conducted PAH measurement using HPLC chromatogram and identified 26 PAHs, which accounted for 49% of the total PAH mass. They assumed that PAH growth was the net effect of acetylene addition to PAH and PAH addition to soot, while soot growth resulted from addition of acetylene and PAH, ignoring oxidation in view of the fuel-rich postflame conditions. Rsoot = Z acet-soot γ acet-soot m acet + Z PAH-soot γ PAH-soot m PAH (1.14) R ΣPAH = Z acet-pah γ acet-pah m acet - Z PAH-soot γ PAH-soot m PAH (1.15) where R = mass growth rate (g cm -3 s -1 ), Z = collision rate calculated from kinetic theory, γ = collision efficiency, and m = molecular mass. Equation (1.14) and (1.15) allow two of the three collision efficiencies to be calculated from the data and known properties. Howard et al. showed γ acet-soot and γ acet-pah in terms of β = γ acet-pah / γ acet-soot, and then computed the fraction of the soot mass growth contributed by PAH addition. The result is shown in Figure 1-6.

35 19 Figure 1-6 Effect of the relative reactivity of soot and ΣPAH. C/O=0.79. ( ) 10mm,( ) 17mm, ( ) 25mm. 41 Figure 1-7 Collision efficiencies at different distances from burner. γ acetylene-soot ( 10 4 ) : ( ) C/O = 0.70, ( ) C/O = 0.79; γ PAH-soot : ( )C/O = 0.70,( ) C/O = Although γ acet-soot and γ acet-pah were not known, Howard et al. assumed that their ratio could be approximated closely enough. According to the collision efficiencies provided in Figure 1-7, Howard et al. claimed that the collision efficiency of PAH with soot was 5000 times higher than that of acetylene with soot. Thus, from Figure 1-6, Howard et al. concluded that most (95% or more) of the soot growth occurred by PAH addition, which is totally different from the acetylene-soot growth mechanism proposed by Harris and Weiner. 36

36 20 Frenklach made a comment on this conflict. He stated that the difference between the conclusion of these two groups comes from the assumption made by Howard et al. about the collision efficiencies γ acet-soot and γ acet-pah. Frenklach stated that chemical reactions at aromatic sites of gaseous PAH were much slower, due to the reaction reversibility characteristic of such systems, than those at aromatic sites of solid soot particles, whose larger sizes removed the otherwise rate-limiting dependence on reversible ring cyclization. However, Frenklach did agree that PAH played an important role in soot formation. The problem is how to identify the relative contribution to soot growth from both acetylene and PAH. Frenklach claimed that the surface growth of soot mass was primarily determined by two processes: acetylene addition via the HACA reaction sequence and PAH condensation on the particle surface. He thought the relative contribution of each of these processes appeared to change with experimental conditions Soot Oxidation The final soot amount produced in the combustion environment depends not only on the process of soot growth, but also on the oxidation process. In fact, the oxidation process is parallel with soot growth and happens during the entire soot formation process. Potential soot oxidants include O 2, CO 2, and OH.

37 21 Frenklach stated that the major oxidation process occurs at the very beginning of soot particle growth, which is the soot particle nucleation period, where a rapidly decreasing concentration of O 2 in fuel-rich environments is observed. 43 According to Neoh et al., 44 the hydroxyl radical is known to be the most abundant oxidizing species under fuel-rich condition. Since OH could suppress soot formation via oxidative destruction of precursors, they stated that OH concentration might be an important factor in soot precursor kinetics. Lucht et al. 45 confirmed that OH is the limiting oxidative reactant under fuel rich condition and that soot field tends to decrease with an increase in OH concentration. Experimental studies conducted by Liu et al. 46 showed that CO 2 has chemical effects on soot formation reduction, in addition to dilution and thermal effects. They suggested that the chemical mechanism of CO 2 addition might be to promote the concentrations of oxygen atom and hydroxyl that in return increase the oxidation of soot precursors in soot formation regions. An experimental study conducted by Vandooren et al. 47 on CO 2 addition to rich but non-sooting CH 4 /O 2 /Ar premixed flames showed that reaction CO 2 +H CO+OH is responsible for the promoted concentration of hydroxyl. They also observed that the concentration of acetylene, the dominant soot precursor species, decreases as a result of CO 2 addition. 1.4 General Soot Physical Properties Because of the development of methods of thermophoretic sampling and analysis by transmission electron microscopy (see Figure 1-3), the understanding of soot

38 22 morphology in flames has been greatly extended in recent years. After detailed measurements, Urban et al. 48 stated that soot particles consist of nearly spherical primary particles with relatively uniform diameters smaller than 60 nm, and whose probability density functions generally satisfied Gaussian distributions. The primary particles tend to be somewhat merged rather than just touching one another. Other important physical properties of soot aggregates include their density, porosity, and composition. Not surprisingly, these properties generally are similar to the properties of carbon blacks, however, there are two major exceptions: soot during the last stages of oxidation has a significant porosity, even extending to the presence of hollow cenospheres, and soot from some internal engine combustion processes contains surprisingly large levels of volatile matter. Otherwise, soot densities are typical of carbon blacks, with values in the range of kg/m Soot aggregates also appear to be relatively nonporous with quantitative (BET) measurements of surface area compatible with shapes observed on TEM photographs except as noted earlier. Soot also mainly consists of carbon except for the small primary particles of recently nucleated soot. For example, soot emitted from long residence time buoyant turbulent diffusion flames (involving the combustion of toluene, benzene, acetylene, propylene and propane burning in air) had the following elemental mole ratio ranges: C/H of , C/O of and C/N of The uniformity of primary soot particles is also an important issue because it is an indication of fundamental changes happened to soot precursor during the soot inception process. Thus, a number of researchers have undertaken soot primary particle

39 23 microstructure studies using High Resolution Transmission Electron Microscopy (HRTEM), see Lahaye and Prado, 50 Dobbins et al.. 49 These studies generally show that primary soot particles contain a variety of internal structures depending on the particular regions where the primary particles are formed. The most common variations of the primary particle structures happen near the core (corresponding to particle nucleation period) and also near the surface of the primary particles. This behavior is illustrated by Figure 1-8, which is the sketch of soot emitted from a diesel engine due to Ishiguro et al.. 51 Ishiguro et al. indicated that the structure near the center of the primary soot particles resulted from coalescence of large PAH molecules to form a nucleation site, which was very different from the more layered and regular structure resulting from soot particle. Figure 1-8 Sketch of the structure of soot emitted from a diesel engine. 51

40 Soot Reduction by Additives The amount of soot produced can be affected by introduction of a variety of additives into flames, such as inert diluents, gaseous additives, metallic additives, and oxygenated additives. An additive in general affects soot formation through three effects: (1) Thermal effect caused by change in flames temperature; (2) Dilution effect as a result of the concentration reduction of reactive species; (3) Direct chemical effect due to participation of additive in soot formation and oxidation reactions. In reality, these three effects occur simultaneously and are intimately coupled. For example, introducing an inert additive not only reduces the concentrations of reactive species (dilution effect), but also lowers the flame temperature (thermal effect). The direct participation of an additive in chemical reactions in general also alters the flame temperature. This variation of flame temperature should be considered as part of the chemical effect of additive besides its thermal effect. Even though it is hard to isolate these three effects precisely, for the convenience of analysis, many researchers are still trying to separate the effects of additive on soot reduction. Thus, in following sections, soot reduction mechanisms by various additives are discussed according to their relative dominant effects, i.e. thermal, dilution, and chemical effects.

41 Inert Diluents It was observed that addition of inert diluents such as N 2 and Ar could reduce soot production in both premixed and diffusion flames. 52,53,54 Schug et al. 54 attributed this reduction in soot to lower flame temperature because this effect could significantly decrease the nucleation and growth processes. However, Kent and Wagner 55 argued that the amount of soot reduction through fuel dilution was more than what can be explained solely in terms of flame temperature, and the fuel concentration could be a significant factor influencing the soot production rate. In view of the above considerations, Axelbaum 56 et al. tried to identify the individual effect of flame temperature and fuel concentration. To isolate the effects of dilution and temperature in a systematic way, they substituted a given volume flow rate of nitrogen on the oxidizer side by an equal volume flow rate of argon which allowed the maximum flame temperature to be increased without affecting the concentrations of the reactive species. This feasibility is attributed to the different specific heats of argon and nitrogen. Based on their experimental work on counterflow diffusion flame, they concluded that nitrogen addition to ethylene resulted in 64% reduction of the soot as a consequence of both dilution and temperature effects, among which 55% reduction was attributed to the dilution effect and the remaining 9% to the temperature reduction.

42 Gaseous Additives Haynes et al. 57 investigated the influence of various gaseous additives on the overall process of soot formation in a premixed ethylene/air flames. Their results are shown in Figure 1-9 and Figure Figure 1-9 Influence of various gaseous additives on the critical C/O-ratio for the onset of sooting in an ethylene/air flame. 57 Figure 1-10 Influence of gaseous additives on the downstream ( t 30 ms ) soot yield in an ethylene/air flame. C/O ~ 0.76, fv~ 1.5 x

43 27 Figure 1-9 shows the effect of additives on the critical C/O-ratio, which corresponds to the appearance of soot, and Figure 1-10 shows the influence of gaseous additives on the downstream (t 30 ms) soot yield. Based on their experimental results, Haynes and Wagner concluded that: (1) At lower concentration level (<1 mol. %), H 2 O could be treated as an inert diluent, similar to N 2 and Ar, since they did not influence the critical C/O ratio and had no effect on the amount of soot. (2) Addition of ammonia (NH 3 ) increased the critical C/O ratio for the onset of soot and reduced soot production in richer mixtures. Haynes et al. suggested that in fuel rich flames NH 3 was converted to HCN, which was actually inert in the absence of oxidants. (3) For NO, there was a threshold concentration below which neither the critical C/O ratio nor the soot yields was affected. Haynes et al. attributed this threshold concentration to two effects: one is the conversion of NO to HCN as the case of NH 3, the other is that NO catalyzes the radical equilibrium in reaction H + OH H 2 O and could significantly alter the post-flame conditions. (4) H 2 S, SO 2 and SO 3 could substantially reduce the amount of soot produced. Haynes et al. attributed the effect of these sulphur-containing additives to a reduction in the amount of considerable soot-forming materials produced in the reaction zone, and claimed that their relative effects were consistent with the trend of increasing effectiveness with increasing oxygen content.

44 28 Also Lawton 58 indicated that effects of SO 2 addition on PAH concentrations and soot particle characteristics in fuel-rich premixed flames were similar to effects observed for oxygen addition to such flames. The amount of carbon removed from the soot was consistent with the significant consumption of SO 2 and the resultant formation of species such as CO, CO 2, CO, S, and CS 2 in the sooting zone. Therefore, Lawton concluded that the reduction in the soot volume fraction of SO 2 could be attributed to increased oxidation of, either soot particles or gaseous species that contributed to soot growth. Failing to observe a decrease in number density of soot particles after introduction of above gaseous additives, Haynes et al. stated that the gaseous additives mentioned above did not affect coagulation, which is different from the ability of ionized metal additives such as Na, K, and Cs to inhibit strongly this process. Therefore, they concluded that the action mode of the gaseous additives was not ionic Metallic Additives Howard and Kausch 59 made an overview of the soot reduction effect of metallic additives and concluded three mechanisms by which additives seem to function in flames. Mechanism I (Na, K, Cs, Ba) This is an ionic mechanism in metallic additives ionize extensively in the flame. The additive ions act on natural flame ions (both molecular and particulate) to decrease the soot precursor nucleation or coagulation rate. The result is a decrease of the amount of soot formed, or a shift of particle size

45 29 distribution to smaller sizes which burn out more quickly. Bulewicz et al. 60 claimed that easily ionizable additives, such as Na, K, Cs, and Ba, could neutralize flame ions and, therefore, inhibit rapid formation of soot precursor via flame ion-fuel molecule chain reaction. Alternatively, some researchers proposed that the additives altered the electrical charge on small soot particles, which inhibited their growth by agglomeration. Thus, these small soot particles were more susceptible to oxidative attack. 61,62 However, Mitchell and Miller 63 found that an increase in free-electron content caused by burning magnesium in the flame was observed to have little effect on the formation of soot. Thus they suspected the validity of the ionic, as these electrons would recombine readily with the flame ions. Mechanism II (Ba, Ca, Sr) Additives in this mechanism undergo a homogeneous reaction with flame gases to produce hydroxyl radicals which rapidly remove soot or gaseous hydrocarbon soot precursors. This action appears to occur throughout the flame, with significant decreases in flame radiation in the early flame zones. 64 Mechanism III (Mn, Fe, Co, Ni) This mechanism, which only occurs to appreciable extents late in the flame (oxygen-rich secondary zone), is an acceleration of the oxidation rate, possibly by occlusion of the metal in the soot particle. No significant decrease in primary zone flame radiation is typically observed with this mechanism.

46 Oxygenated Additives The idea of using oxygenated fuels as a means of producing cleaner-burning diesel engines was introduced over 50 years ago. Consequently, a large number of oxygenates in the form of carbonates, ethers, esters and alcohols have been added to diesel fuels. 65 Miyamoto et al. 66 investigated four kinds of oxygenated agents added to diesel fuels in a DI diesel engine. Figure 1-11 is their results showing smoke and particulate emissions with oxygen content in the blended fuel on the abscissa. They found significant smoke and particulate reduction with every oxygenated agent, and emissions decreased linearly with increasing oxygen content in the fuels. Because the dispersion among the kinds of oxygenated agent is small, they claimed that the degree of smoke and particulate suppression almost only depends on the oxygen content in the blended fuel regardless of the kinds of oxygenated agent. Their further combustion analysis with the twodimensional two color method showed that, during the combustion process, soot concentration in the flame decreased with the addition of oxygenated agent while the flame temperature distribution was almost unchanged. Thus, Miyamoto et al. suggested that oxygenate agent did not have a large effect on the entire combustion but tended to supply more oxygen to the rich mixture region resulting in low smoke emission.

47 31 Figure 1-11 Smoke emissions from diesel engines with oxygenated additives, from Miyamoto et al. 66 However, some researchers observed that specific types of oxygenates may be more beneficial in particle emission reduction, e.g. the use of ethers versus alcohols. Liotta and Montalvo 65 tested several oxygenated compounds as diesel fuel addictives and concluded that the type of additive (ether or alcohol) influenced the effect on the particulate reduction. As a summary of literatures, the following factors are found to be responsible for the soot suppression effect of oxygenated additives: temperature, carbon-carbon bond, oxygen content, molecular structure, and hydrogen effect.

48 32 The addition of oxygenated additives to the fuel stream may change the flame s temperature because of difference in heats of combustion and effects on gas composition in the flame. It is well known that the processes of soot nucleation and growth, and the pyrolysis of fuel and additive are all sensitive to temperature. During an investigation of DMC and ethanol as fuel additives in a diesel engine, Shih 67 blended diesel fuel with 5% and 10% by volume DMC and 10% and 20% by volume ethanol and found that both DMC and ethanol reduced the exhaust gas temperature as well as smoke opacity. Instead of thermal effect, Rubino and Thomson emphasized the importance of the carbon-carbon bond in the fuel stream on soot production. 68 Different fuel molecules may contain different numbers of C-C bonds, for instance, propane has two C-C bonds, ethanol has one, and DMC none. Rubino and Thomson showed that soot production decreased as the number of C-C bonds of the fuel stream decreased and claimed that fuel without C-C bonds could not produce significant levels of the species such as acetylene, ethylene or unsaturated radicals, which are known to lead to aromatic species. Another effect of oxygenated additive addition is the influence of fuel oxygen content on soot production. The fuel oxygen content is known to be an important factor for soot reduction. 69 Beatrice et al. explored the influence of fuel oxygen content on acetylene level and found that the introduction of oxygen content in fuel may produce lower acetylene concentration. Figure 1-12 shows acetylene evolution in the sampling control volume during combustion for several tested fuels in a single cylinder direct injection diesel engine. At about the same cetane number and spray evolution, the acetylene concentration is very well related to the oxygen fuel content, with a decrease of

49 33 the whole curve in butylether with respect to n-tetradecane. When fuel oxygen content increases and volatility decreases, as in the case of monoglyme, a strong drop of the whole curve is detected. Also, these results are congruent with thermal cracking analysis obtained by Fujiwara et al. 70 who found that oxygen in the fuel spray reduces pyrolysis products. Based on the exhaust soot emissions index reported on Figure 1-12, the authors claimed that at the same cetane level and with different oxygen content the emissions of smoke are related to the acetylene peak concentrations. Moreover the relative scaling of acetylene concentration between the fuels agrees with the scaling between the same fuels in terms of soot volume fraction evolution reported by Beatrice et al. 71 Beatrice et al. explained that the pyrolysis of additive generated O and OH radicals, which could attack acetylene and other soot precursors. Figure 1-12 Acetylene concentration evolution of fuels with about same centane number. In the figure, intervals of confidence of 80% are indicated. Test condition: engine speed = 1250 rpm, 4.5 bar of i.m.e.p., start of combustion = TDC. 69

50 34 Curran et al. 72 made another hypothesis for the role that the oxygen content plays in soot reduction. He stated that when oxygenated hydrocarbons were added to fuel, the oxygen remained permanently connected to a carbon atom. No elementary reaction was able to break this bond, so that the carbon atom was unable to participate in any of the reactions of small unsaturated species that have been identified as leading to aromatic compounds and soot. Kitamura et al. 73 found that there was a remarkable difference of small unsaturated species levels between DME and ethanol, which had the same molecular formula of C 2 H 6 O but different molecular structure. Therefore, they suggested that the potentiality of oxygenated fuel on soot precursor formation was dominated by molecular structure, as well as by fuel oxygen content. Frenklach and Yuan 74 found that the presence of oxygen primarily influenced the small molecule reactions during the induction period. They investigated the sooting tendencies of alcohol (methanol and ethanol) and benzene mixtures and noted that the suppression of soot formation was due to the ability of oxygenated blends to increase induction time and decrease the rate and amount of soot formed. Additionally, Frenklach found that the decomposition of ethanol effectively decreased the degree of hydrogen atom equilibrium as compared with methanol, thus, allowing ethanol to have a stronger suppression effect on soot formation from benzene. According to Frenklach s HACA mechanism, hydrogen atoms are particularly important in soot formation, since hydrogen atoms reactivate aromatic molecules, and, thereby, propagate the ring growth processes.

51 35 Marinov et al. 75 pointed out that a hydrogen-atom pool was established by reaction HCO + M H + CO + M (1.16) where M is the usual third body. This reaction consumes propargyl (H 2 CCCH) radical and benzene by hydrogen atom abstraction reactions therefore limiting the growth of benzene.

52 36 Chapter 2 PRINCIPLES OF DIAGNOSTIC METHODS Experimental studies of soot suppression effect of additives in flame environments strongly rely on suitable diagnostic techniques, due to the complexity of these phenomena. The use of laser diagnostics has been proven to be efficient in the investigation of sooting processes because of their fast response, high sensitivity, and non-intrusive character. There are many laser diagnostic techniques being used. In the present project, however, the main optical flame diagnostic methods are emphasized on Laser-induced Incandescence (LII), Laser Extinction (LE), and Laser-induced Fluorescence (LIF). Laser-induced Incandescence is used to measure soot volume fraction and investigate the effect of additives on soot reduction; Laser Extinction is used to calibrate LII signal and provide quantitative value of soot volume fraction; Laser-induced Fluorescence is used to detect concentrations of PAH, the major soot precursor for soot formation. In following sections, an introduction of these diagnostics is given. As a main and relatively new diagnostic technique in this research program, Laser-induced Incandescence is introduced in more details than Laser Extinction and Laser-induced Fluorescence. Complete descriptions of Laser Extinction and Laser-induced Fluorescence

53 37 can be found in books by Van de Hulst, 76 Kerker, 77 Bohren and Huffman, 78 and Eckbreth. 79 At the end, temperature measurement by thermocouple is also introduced. Due to the important role of temperature throughout the soot formation and modeling study, temperature measurement was conducted in present study. Thermocouple method was chosen since it is widely used for the measurement of gas temperatures in flames and combustion environments. 2.1 Laser-induced Incandescence Because of its conceptual simplicity and ease of implementation, Laser-induced Incandescence (LII) is currently preferred for determining particle volume fraction in highly turbulent reacting flow fields where the particles are optically absorbing. The theory for LII has been placed on a sound basis and the remaining issues regarding particle optical and physical properties have been clearly identified in contemporary published work. LII was first noticed by Eckbreth in 1977, 80 when it produced an interference signal in coherent anti-stokes Raman scattering experiments in particle-laden flames. Subsequent work by Melton 81 and Dasch 82,83,84 placed this technique on reasonably firm theoretical ground regarding the potential of LII for quantitative soot volume fraction and particle size measurements. Application of LII as a soot diagnostic for qualitative

54 38 measurements in transient flowfields followed somewhat later with the work by Dec and co-workers as part of their investigations of soot formation in diesel engines. 85,86 Shortly thereafter, several research groups published papers addressing quantitative measurement approaches for soot volume fraction. 87,88,89,90 Recently, Santoro and Shaddix wrote a detailed review article about LII, in which basic theory and important considerations for experimental setup and calibration of a LII measurement system are discussed. 91 Most of following work about LII refers to this article Theoretical Analysis LII is generated when a cloud of soot is irradiated with an intense laser light. The soot particles are heated to a temperature well above the surrounding gas temperature due to the absorption of laser energy. The heated soot particles subsequently emit black body radiation corresponding to the elevated soot particle temperature. If the laser fluences are high enough, the vaporization of small carbon fragments, typically C 2 and C 3, from the surfaces will take place once the vaporization temperature of carbon of 4000 K is reached.

55 39 Hofeldt built a computational model, which describes the heat transfer between particle and its surroundings, as well as the interaction of particle with the incident laser radiation. 92 In this model, the energy balance is represented by Equation (2.1): m s d( cst dt s ) H M v v dm dt s = qc abs ha ( T s s T ) CabsEb, λ ( Ts ) dλ + Cabsε b, λ Eb, λ ( Ts ) 0 0 (2.1) dλ where: ms = mass of a particle; cs = specific heat of the particle; Ts = particle temperature; T = gas temperature; Hv = enthalpy of vaporization; Mv = molecular weight of vapor; Q = laser excitation intensity; Cabs = particle absorption cross section; H = convective coefficient; λ = wavelength; E b,λ = blackbody spectral irradiance; ε b,λ = spectral emissivity; The left-hand side terms are: 1) the rate of increase of energy stored in the particle and 2) the loss in particle energy due to vaporization. The right hand side terms are: 1) the rate of energy absorption from the laser pulse, 2) the collisional cooling rate, which is written in terms of convection, 3) radiative emission, and 4) radiative absorption. Hofeldt claimed that the last term can usually be ignored unless there are significant contributions from the surroundings as would be the case if very high temperature walls are present.

56 40 As a black body, the soot particle radiates according to Planck s equation. LII signal, S, collected over a given solid angle, Ω, from a distribution of soot particles, can be written as τ Ω S = Vmv Cn ( t) W ( t) N( Ds ) CabsEb, λ ( Ts ) dλ dds dt (2.2) 4π where V mv = measurement volume; C n (t) = soot particle number density; W(t) = windowing function, which is the signal collection gate whose duration is τ; N(D s ) = normalized size distribution. Santoro et al. stated that a critical aspect in the successful application of LII to absorbing particles lies in the fact that particles, either in terms of their equivalent sphere diameter or primary particle diameter lie in the Rayleigh scattering regime, that is λ/d < 0.3. Here λ is the wavelength of the incident laser beam. In response to an idealized laser pulse based on the black body radiation laws and the soot particle energy balance, Melton 93 showed that Equation (2.2) turned out to be Cn t) W ( t) N( Ds ) x S = Ccal ( Ds dλ dds dt (2.3) where C cal = calibration factor; x = (3+0.15λ -1 ) ; Melton pointed out that LII signal S is proportional to D s x, and, therefore, provides a measurement that corresponds closely to the volume fraction of absorbing particles. This relationship forms the basis of using LII for soot volume fraction measurement.

57 41 According to the definition, x = (3+0.15λ -1 ), longer wavelength incident light would make Melton s parameter x approach a value closer to 3, and, at the same time, Rayleigh scattering criteria would be better met. However, Shaddix and Smyth 94 argued that for particles whose sizes exceed the Rayleigh limit, they may emit on less than a volumetric basis. In addition, considering the uncertainty associated with descriptions of particle heating, and laser interactions with vaporizing material, Shaddix and Smyth suggested that the expression developed by Melton should only be considered a rough approximation Experimental Considerations A typical LII experimental setup is shown in Figure 2-1. It consists of a highenergy pulse laser, focusing optics, collection optics, an appropriate optical filter, a photodetector, and a suitable data acquisition system. LII can be applied either as a point, line, or two-dimensional laser sheet measurement. The essential elements of the experimental setup are similar in each case with only the focusing and collection optics differing to meet the particular measurement desired.

58 42 Figure 2-1 Optical setup for 2-D LII measurements. LDF laminar diffusion flame, BB beam block, DM dichroic mirror, ½-λ - halfwave plate, GL Glan-Laser prism, A aperture, P prism, CL cylindrical lens, SL spherical lens, and ICCD intensified CCD. 91 Even though the apparatus for implementing LII technique is quite simple in principle, Santoro et al. mentioned that there are several critical parameters needed to be taken care of, for example, the laser excitation energy and wavelength, laser intensity profile, spectral detection region, and detection gate width and timing. In following paragraphs, these parameters will be considered in terms of their impact on LII measurements.

59 Laser Excitation: Intensity Profile, Energy, and Wavelength For a diagnostic application, the effect of the laser intensity profile is important and should be characterized for each system in order that LII signals can be properly interpreted. Many researchers have shown that different laser intensity profiles can result in distinct behaviors of the LII as a function of laser fluence. For Gaussian beams the LII signal often exhibits the behavior evident in Figure 2-2, where the signal in the plateau region actually increases moderately with increased laser fluence. 94 Quay et al. 95 has shown that for a pulse Nd-YAG laser at 532 nm, a minimum laser fluence of about 0.2 J/cm 2 is required to achieve a saturation or plateau condition with respect to the incident laser energy fluence. In contrast, Vander Wal and Jensen reported that the signal they observed achieved a plateau region between 0.2 and 0.4 J/cm 2 and then decreased as laser fluence increased. Work by Ni et al. 96 and Witze et al. 97 using a uniform beam profile have shown that with increasing laser fluence the LII signal achieves a maximum and then decreases to a plateau at higher fluences.

60 44 Figure 2-2 Fluence dependence of LII measured in steady laminar diffusion flames. Data were collected at H = 20 mm in the ethylenec/air flame for detection gate durations of 19 ns (+) and 85 ns ( ) both gates opening coincident with the arrival of the 5 ns laser pulse. Data are also shown for the methane/air flame at H = 50 mm with 85 ns gate ( ). Raw signals for each condition have been normalized to a value of 1.0 at a fluence of 0.6 J/cm 2. The solid line shown is the least-squares power-law fit of the methane data for fluences greater than 0.03 J/cm 2 ; the fit follows the expression signal fluence reference [94].

61 45 Santoro et al. thought a decreasing trend in signal intensity is consistent with increased mass vaporization from the particle, since vaporization results in less soot mass in the probe volume, resulting in smaller LII signals. In the case of the Gaussian beam, Santoro et al. explained that, as the laser fluence increases the region of the beam cross section that exceeds the LII signal threshold value increases. The resultant larger sampling volume compensates for the signal losses due to vaporization of the particles near the center of the Gaussian profile. The relative effects of the mass vaporization and the increasing probe volume are determined by the specific characteristics of the laser intensity profile and the transmission optics employed to form the optical excitation region. With respect to the selection of the wavelength for LII excitation, a wide range of wavelengths can be employed as long as the particles exhibit sufficient absorption at that wavelength. In general the selection of the laser wavelength involves minimizing potential interferences that can accompany the laser excitation process such as laserinduced fluorescence from other species present. Santoro et al. stated that in the case of soot diagnostics this interference usually means LIF from polycyclic aromatic hydrocarbon (PAH) species, which absorb and fluoresce in a broad region throughout the UV and visible regions of the spectrum. For example, LIF from PAH can be excited by the second harmonic of Nd-YAG laser at 532 nm, but will not be excited by the fundamental laser line at 1064 nm. Thus, the use of 1064-nm laser line is often cited as an approach to eliminate this source of interference. 98, 94, 99 However, the use of an infrared

62 46 excitation source does introduce some inconvenience since the beam is not visible and alignment is more tedious. Vander Wal et al. 100, 101 found that the LII excitation process is an intrusive one in which the particles in the probe volume are subject to high heating rates that can affect the physical and optical properties of the particles. Santoro et al. stated that although such effects do not appear to limit the quantitative capability of LII to measure volume fraction, it does raise concerns when modeling the process as the physical and optical properties are varying in a manner usually not known. In the case of soot particles where the optical properties are already somewhat uncertain, this only further complicates the theoretical problem Spectral Detection Region Because of the black body nature of incandescence, the spectral region that can be used to detect LII signal is quite wide. However, care must be exercised in the selection of the spectral bandwidth over which measurements are obtained. Shaddix et al. 94 found that contributions from C2 emission generated by the laser heating process can introduce interferences particularly in the wavelength region between 420 and 620 nm. C2 interferences are particularly important in flames where the soot concentration is large (> 2 ppm) and when a frequency doubled YAG laser (532nm) is used to excite the LII. They claimed that the use of narrow band filters, which have peak

63 47 transmissions wavelengths at 450 nm or shorter wavelengths, can effectively reduce C2 emission interferences. Furthermore, Santoro et al. claimed that since interferences resulting from fluorescence due to the presence of PAH species generally are red-shifted with respect to the excitation laser, the approaches that address the C2 interference problem can typically resolve the PAH interference, unless UV laser sources are used to excite the LII signal. Based on above statements, and, also, noticing that the natural luminous soot radiation decreases sharply towards the UV, Santoro et al. recommended shorter detection wavelengths to be used Detection Gate Width and Timing Figure 2-1 shows typical time-resolved LII signals obtained for soot particles formed in a laminar diffusion flame. 96 The decay time is on the order of 100 to 200 ns as compared with the excitation period, which corresponds to the laser pulse duration of 10 ns. LII signals vary with time and thus require some selection of the period over which measurements will be made and when to start the measurement period. In general, researchers have used two approaches: (1) prompt gating and (2) delayed gating. In the first approach a short gate width (10 to 50 ns) is promptly timed to extend from the onset of the laser pulse 102, 94 or very shortly after the laser pulse. 96 The major advantage of this approach is that it minimizes effects of particle size differences that arise when delayed detection times are used, as will be discussed below. One potential

64 48 disadvantage of the prompt detection approach is that interferences from vaporized species or fluorescence from PAH species may also contribute to the signal unless proper spectral filtering is used as discussed above. Figure 2-3 Temporal profile of a LII signal obtained in the etheneair laminar diffusion flame at heights of 10 and 30 mm above the fuel tube exit and at the radial locations corresponding to peak soot volume fraction for these heights. Cignoli et al. 103 recommended delayed detection a means to discriminate against interferences from vaporized species or species that fluoresce from the excitation laser used. They claimed that since fluorescence decay times are much shorter than the LII decay time, delaying the gate can effectively eliminate these sources of interference.

65 49 However, Santoro et al. argued that since the cooling rates of the particles differ, this approach tends to weight larger particle sizes as either the gate delay time or gate duration is increased. Thus, Santoro et al. suggested the use of the prompt gating approach for most applications, unless special experimental conditions indicate a delayed gate has advantages Calibration In many applications it is important to place the LII measurements of soot properties on a quantitative basis through calibration of the signals. Quantitative measurements require a suitable calibration technique that ultimately determines the accuracy of the measurements. According to the review article of Santoro et al., there are presently four approaches used to calibrate LII signals for soot volume fraction, including laser extinction, cavity ringdown, gravimetric measurement, and 3-color pyrometry measurement. These methods are introduced briefly in the following paragraphs Calibration via Laser Extinction Most calibrations of LII have been performed by comparing LII signals to laser extinction measurements, the traditional in situ diagnostic for soot concentrations. 88, 90, 94,

66 50 104, 105, 106 The accuracy of an LII calibration for soot volume fraction based on the light extinction approach is limited by uncertainties concerning the appropriate soot extinction coefficient to use for a given application and by uncertainties over the extent of soot optical property variation within a given combustion or flame system. Santoro et al. stated that recent measurements of the nondimensional extinction coefficient, Ke (= 6πE(m) in the Rayleigh limit), have consistently yielded values between 8 and 10 at visible wavelengths for smoke emitted from both laminar and turbulent flames of various 107, 108 fuels. In contrast, the variety of soot index of refraction measurements conducted over the past three decades tend to cluster around values such as m = i, widely attributed to Dalzell and Sarofim; 109 this value of m gives a Ke of 4.9 in the Rayleigh limit. Another widely quoted value for soot index of refraction, Lee and Tien's i, 110 gives Ke = 3.6 in the Rayleigh limit. Thus, calibration of LII with light extinction for the determination of soot concentrations has an uncertainty on the order of a factor of two in most applications, unless the nondimensional extinction coefficient is directly measured or in some manner known to better accuracy Calibration via Cavity Ringdown However, when soot concentrations are very low, calibration via the traditional extinction methodology or via a gravimetric approach is very difficult. To overcome this problem, Vander Wal 111, 112 explored the use of cavity ringdown (CRD) measurements of laser light extinction from soot as a means of calibrating LII signals for soot

67 51 concentration. He showed that the use of CRD to calibrate flows containing low soot concentrations removes the difficulties associated with extrapolating an extinction calibration down to soot concentrations that may have different aggregate structures and primary particle sizes than the soot particles characteristic of higher concentration flowfields generally used for traditional light extinction calibration. Comparisons between light extinction and CRD measurements show good agreement. With CRD, 70- ppb soot has been calibrated to within a few percent precision with a single laser pulse. As with traditional extinction measurements, the accuracy of a CRD-based calibration is ultimately limited by uncertainties in the optical properties of the probed soot Calibration via Gravimetric Techniques Calibration by a gravimetric technique involves the collection and weighing of soot deposited on a filter in a calibration flowfield from which LII signals are also collected. 113 This approach to LII calibration has the advantage of avoiding the large uncertainties associated with soot optical properties and is the accepted norm for engine exhaust soot measurements, when using a dilution tunnel. In order to convert the measured soot mass to an entrained soot concentration (on a mass basis), the volumetric flowrate through the collection filter and the gas temperature at the point of the LII measurement must be known. To convert the calibration to a volume fraction basis, the mass density of the soot must be either assumed or measured. In addition, it is important that the LII and gravimetric measurements either completely samples the calibration flow

68 52 cross-section or that the soot loading be uniform at the LII and gravimetric sampling points. Finally, the soot collection filter must either be heated during collection or dried later in order to remove water or other condensables from the collected filter mass. This calibration technique is generally limited to those applications in which a significant soot concentrations is produced and, in contrast to the optical calibration techniques, requires filter collection over the course of minutes Calibration via 3-Color Pyrometry Practitioners of laser-induced incandescence have long desired a means of quantifying soot volume fraction directly from the LII signals themselves. Snelling et al. 114 have been the first to attempt to do this. The approach developed by Snelling et al. involves the use of 3-color pyrometry to measure the transient temperatures of the soot particles during the LII laser excitation. This information, coupled with an absolute calibration of the light collection efficiency of the LII detector system (through use of a blackbody, for example) and an assumed value of E(m), allows one to use the measured LII signal strength to solve for the total volume of emitting particles within the field of view of the LII detection system. Dividing the emitting volume by the product of the cross-sectional area of the imaged laser beam/sheet and the beam/sheet thickness, yields the soot volume fraction. Unfortunately, the accuracy of this technique is dependent on the assumed optical properties of the imaged soot during the LII measurement period, and these properties presumably lie somewhere between that of the native soot (with

69 53 unknown properties) and that of graphite. For example, Snelling et al. 114 chose to use E(m) = 0.26, consistent with m = i (a typical refractive index assumed for soot), and measured engine-out soot concentrations with this calibration of LII that were 50% higher, on average, than those measured with gravimetric sampling. Significantly larger values of E(m), such as those determined from the post-flame light extinction measurements of Mulholland and others, would need to be assumed for Snelling's selfcalibrating LII measurements to show good agreement with the gravimetric sampling General Concerns in Calibrating LII Even though each calibration method has its limitation in accuracy, there are some general concerns in calibrating LII. Bryce et al. found that variations in flame temperature and chemical composition affect the heat conduction rate of the gases surrounding the laser-heated soot. 115 Santoro et al. suggested that the calibration is desirable to be conducted at the same temperature, chemical compositions, pressure, and characteristic primary particle size between the calibration source and the LII measurement application. They also claimed that performing prompt LII signal detection was preferred in order to decrease these effects.

70 Laser Extinction Laser extinction (LE) has been a popular optical diagnostic method in soot concentration measurements owing to its simplicity and accuracy. Also, a combination of light scattering and extinction can be used to measure soot particle size. When a cloud of particles is placed in a beam of light radiation, as shown in Figure 2-4, the power detected will be reduced. In this case, the presence of particles has resulted in the extinction of the incident light. The attenuation of light is caused by both scattering and absorption effects of particles. Incident Scattered Detector Figure 2-4 Extinction by a collection of particles. 116

71 55 Beer s law: The light intensity after passing through the particle cloud is given by Lambert- L I = I K dl 0 exp( ext ) (2.4) 0 where I 0 = incident light intensity (W m -2 ); K ext = extinction coefficient for a cloud of particles (m -1 ); L = the path length of the light beam going through particle cloud (m). From light scattering theory, the extinction coefficient for a cloud of particles is related to the number, size and properties of individual particles by K ext π = C 4 n 0 Q ext N( D) D 2 dd (2.5) where C n = the number density (m -3 ) Q ext = the extinction efficiency of a particle; D = the particle diameter (µm); N(D) = normalized particle size distribution function. The value of the extinction efficiency of each particle, Q ext, is a complex function of the size and refractive index of the particle. The exact expression for Q ext does not exist, but various approximate techniques can lead to simple expressions for calculation. One of the most common approximate methods is Rayleigh scattering, which applies to small particles with x<<1 and x m-1 <<1, where m is the complex refractive index and x = πd/λ is the particle size parameter.

72 56 Since soot particles normally are smaller than 60 nm in diameter, and the shape of the individual particle is roughly spherical, Rayleigh approximation may be applied. Strictly speaking, soot particles form non-spherical agglomerates, which take the form of branched chains. 116 So results should be considered in that light. In addition, because the composition of soot varies, the refractive index is uncertain. This leads to error in the calculation of scattering properties and the analysis of experimental results. obtained: Applying the Rayleigh scattering approximation, the extinction efficiency is 2 2 m m 1 Q ext = 4 x Im( ) + x Re( ) (2.6) 2 2 m m + 2 The first term is the absorption efficiency, Q abs, and the second the scattering efficiency, Q scat. Thus, Rayleigh scattering efficiency is proportional to x 4 whereas the absorption efficiency is proportional to x. When visible light with wavelength ranging between 0.4 µm and 0.7 µm is used, according to the definition x = πd/λ, soot particle size parameter x lies in the regime x<<1, absorption dominates over scattering, i.e. Q abs >> Q scat., and soot is heavily absorbing. Thus extinction efficiency can be approximated by 2 m 1 Q ext = Q abs = 4 x Im( ) (2.7) 2 m + 2 So, Lambert-Beer s law becomes I 2 2 L m 1 2 = I 0 exp( π Im( ) C ( ) ) 2 n + 2 N D D dd (2.8) λ m 0

73 57 The volumetric density of particles f v (the volume of particles in a unit of space) is related to the number density of particles C n (the number of particles in a unit of space) by f v π = C 6 n 0 N( D) 3 D dd (2.9) It can be obtained from the extinction measurement by f v 2 D32 I = ln( ) (2.10) 3L Q I m 0 where Q m is the mean extinction efficiency for a poly dispersed particle cloud 2 N( D) Q( D) D dd 0 Q m = (2.11) 0 2 N( D) D dd D 32 is known either as volume/surface mean diameter or Sauter mean diameter Q = (2.12) N( D) D dd 2 N( D) D dd Kontani and Gotoh carried out a detailed analysis on Q 117 m. It has been shown that for particle of size parameter x<0.5, such as soot particles, the difference between Q m and Q(D 32 ) (the mean extinction efficiency based on D 32 ) is less than 5% regardless of the size distribution function. The error will be increased to 10% if the size parameter approaches 1.

74 58 With Q m replaced with Q(D 32 ) and an expression for Q(D 32 ) (Equation 2.7) the volumetric density of particle can be found: f v λ = m 6π L Im m I ln( I 0 2 ) (2.13) It is obvious that the particle volume concentration can be calculated from the ratio of I/I 0 without any knowledge of particle size distributions. The only approximation will be the value of the refractive index to be used. This could be the largest source of error as the refractive index varies with wavelength. Chang and Charalampopoulos 118 have measured experimentally the variation of refractive index with wavelength, which is shown in Table 2-1. Wavelength Refractive index λ(µm) m = n - ik i i i i i0.59 Table 2-1 Refractive indices from Chang and Charalampopoulos 118

75 59 A simple calculation using Equation (2.13) shows that the soot volume concentration is 33% lower if the refractive index is changed from m=1.9-i0.55 to m=1.4- i0.55. Thus the choice of the value of the refractive index has a significant effect on the measured value of soot concentration. 2.3 Laser-induced Fluorescence Laser-induced Fluorescence (LIF) is a well-established technique for detecting the population densities of molecular or atomic species in specific quantum states. In combustion applications this information can be used to determine relevant quantities such as mole fractions, density, temperature, and velocity. 119, 120, 121 LIF is the spontaneous isotropic light emission of molecules that have been selectively driven onto an excited electronic state by tuned laser excitation (optical pumping), then relax to their ground state. The fluorescence power is then directly proportional to the excited state population through the Einstein probability coefficient A for spontaneous emission. In hot reacting media, collisions and chemical reactions can also populate excited states, but the excited populations and the subsequent emissions induced by these processes are much lower than those induced by laser pumping. Absorption of the laser photons by molecules is directly responsible for the population of the excited state in the laser field. Besides relaxation by spontaneous emission of fluorescence at rate A, other depopulation processes such as stimulated emission, collisional quenching, energy transfers or predissociations (at global rate Q in a

76 60 simplified two-level schema) are competitively involved in the interaction. The dynamics of the population transfer must be carefully examined to obtain the concentration of investigated species in the excited state as a function of its global concentration. Then it is made possible to derive that global concentration from the measured intensity of the laser induced fluorescence emission. Following paragraphs will emphasize the simple regimes for which the fluorescence emission is locally proportional to the laser irradiance and to the molecular population in its lower state. Different calibration procedures may be used to determine the proportionality coefficient in order to derive absolute concentration data from measured fluorescence intensity. In many cases a simple reference sample of the investigated molecular species is not currently available and the calibration must be performed in a particular zone of a reacting flow where the absolute concentration of the molecule has been calculated or can be measured by another technique such as absorption spectroscopy Theory of Laser-induced Fluorescence Initial theoretical work on fluorescence was carried out by Piepmeier 122 to describe the molecular dynamics of fluorescence experiments of atomic species seeded into analyzer flames. This was achieved with a rate equation analysis of an ideal two-

77 61 level system by assuming that the populations of these levels reach a steady state. The following sections are a brief account of the rate equation analysis. Upper Level 2 I v B 12 I v B 21 Q 21 A 21 Lower Level 1 Figure 2-5 Schematic diagram of two-level model of induced fluorescence. In the two-level model of fluorescence, one considers only two molecular quantum states that are directly populated or depopulated through interaction with the laser light. Transfer of energy resulting in the population of neighboring quantum states is neglected. The energy transitions and the transfer mechanisms that are considered in this model are summarized in Figure 2-5. Each mechanism is represented by a rate (s -1 ) and a direction. The rates of stimulated emission and absorption of photons resulting from laser interaction are designated by I v B 21 and I v B 12, respectively, where I v is the laser spectral intensity [J/(cm 2 s Hz)] and B 21 and B 12 are the Einstein B coefficients for the transition (cm 2 Hz/J). Spontaneous light emission from the upper energy level is described by the Einstein A coefficient A 21 (s -1 ), and the collision quenching rate from the upper level to lower level is denoted by the term Q 21 (s -1 ). The laser spectral bandwidth is assumed to be

78 62 larger than the molecular absorption linewidth so that there is a complete overlap, rendering the details of the absorption lineshape irrelevant. In current application of laserinduced fluorescence, the temporal dynamics of the excitation process are not resolved, and we use average intensities, I τ τ = I ( t) dt v v (2.14) 0 where τ is the laser pulse duration. In typical flame environments, the upper-state lifetime is on the order of 10-9 s whereas the laser duration is about 10-8 s, and in this case one can also use average population densities, τ 2 τ = N2( t) dt N (2.15) 0 Stimulated emission from molecules transitioning from the upper level to the lower level possesses the same momentum and phase as the incident laser radiation. Spontaneous emission, however, has random momentum and phase and is emitted into 4π steradians. It is portion of this radiation that is collected and constitutes the fluorescence signal. The fluorescence signal can be described by the equation = τη Ω N 2 A V (2.16) 4π S t 21 where η is the efficiency of the collection optics, which collect photons through a solid angle Ω; N 2 represents the number density of molecules in the upper state due to laser excitation; and V is the collection volume imaged onto one detector pixel. The collection

79 63 volume is defined by the thickness of the laser sheet multiplied by the area of the sheet imaged onto a single pixel. The rate equation describing the population in the upper state may be written as dn dt = N I B N ( I B + Q ) 21 (2.17) 2 A 1 v 12 2 v The first term on the right represents the rate of population transfer from the lower state to the upper state due to stimulated absorption, and the remaining terms represent depopulating mechanisms; stimulated emission, collisional quenching, and spontaneous emission, respectively. In situations where the duration of the laser pulse is long compared with the quenching time of collision, it may be assumed that the system reaches a steady state; thus Equation (2.17) becomes N 1Iv B12 2 v A = N ( I B + Q ) 21 (2.18) In flames where the temperature seldom exceeds 3000 K, the upper energy level state is initially empty. The steady state populations then satisfy the constraint where N + = (2.19) 0 1 N2 N1 0 N1 is the initial population in the lower state. Substituting Equation (2.18) into Equation (2.19) and introducing the result into Equation (2.16) leads to S f = Ω VN I B A 0 v τη 1 (2.20) 4π Iv( B12 + B21) + Q21 + A21 The initial population of the lower level is related to the total number density N t of the species being probed by the Boltzmann fraction f B, so that Equation (2.20) becomes

80 64 S f = Ω I B A v τη Vf BNt (2.21) 4π Iv( B12 + B21) + Q21 + A21 This is the basic fluorescence equation, which may be used to relate the measured fluorescence signal to the total number density Nt. A particularly simple form is obtained for weak excitation [i.e., when I v (B 12 +B 21 )<<Q 21 +A 21 ]. In this limit, S f = Ω I B A v τη VfBNt (2.22) 4π Q21 + A21 Since the quenching is much larger than the spontaneous emission probability (Q 21 >>A 21 ) A = η Ω E N (2.23) 21 S f Vf B B12 4π Q21 v t where τi v is replaced by E v, the laser spectral fluence[j(cm 2 Hz)]. The two-level model as presented above is quite appealing because it is simple; however, it does not account for many physical processes that are potentially important in laser-induced fluorescence measurements. In particular, the model neglects the presence of other molecular energy levels that may play a role in the energy transfer processes. This aspect is described in refined theories that take into account energy transfer between the level being directly populated by the laser excitation and nearby rotational and vibrational energy levels. Under certain experimental conditions these additional levels must be included, especially when the laser intensity is great and the weak excitation limit is no longer valid.

81 LIF Calibration In the regime of linear laser excitation, the quasi-steady state fluorescence signal is proportional to laser power and to the local population of investigated molecule in the laser sheet. But it is also inversely proportional to the local rate of electronic quenching which has to be known as a function of different collision partners. Thus calibrations are required to investigate the local influence of this quenching on the concentration measurement. The quenching rate can be calculated in various flames using available data for the specific quenching cross sections and weighting by the local mole fractions of these different collision partners. 123, 124 Another approach is to perform direct determination of the local effective quenching rate in low pressure premixed flames by measuring the decay rate of the time resolved fluorescence signals. However, in a more practical way, a calibration of LIF signal is usually performed under well-known conditions in well-described flames like the McKenna burner, where the absolute concentration of the investigated molecule can be calculated 125 or can be measured by another technique such as line of sight absorption spectroscopy. 126

82 Temperature Measurement In order to avoid the need of modeling heat losses during the use of PREMIX code, a known flame temperature profile is required. There are also computational advantages to using a known temperature profile. The most severe nonlinearities in chemical kinetics come from the exponential dependence of the reaction rates on temperature. Thus, eliminating temperature changes from the iteration makes the flame problem considerably easier to solve. Two methods were proposed to measure the flame temperature profile: one is twocolor, and another is thermocouple. Two-color method utilizes the thermal radiation from soot particles, and directly measures their temperature. Unfortunately, in present flames, the soot concentration during the range of 0 to 5 mm above the burner surface is lower than 1.0E-2 ppm, which is too low to provide accurate temperature measurement result by two-color method for present experimental setup. However, almost all the main reactions happened within the range of 2 to 4 mm, in order to run the simulation program correctly, an accurate temperature profile has to be provided especially for that range. Obviously, two-color method can hardly meet this requirement. After several unsuccessful tries, two-color method was finally given up. Thermocouples are widely used for the measurement of gas temperatures in flames and combustion environments. Generally, correction is needed to accurately determine

83 67 the gas temperature, because in most combustion applications the temperatures are high enough and energy loss from the thermocouple can not be ignored. There are four heat transfer modes during the thermocouple measurement: convection between the gases and the thermocouple, radiant heat transfer between the thermocouple and its surroundings, conduction along the thermocouple wires, and surface-induced catalytic reactions. Among these four modes, convective heat transfer is desirable since the objective of a thermocouple measurement is to determine the local gas temperature. Other heat transfer modes, especially thermal radiation, can cause the difference between the thermocouple measurement result and actual local gas temperature. Rapid thermal conduction along thermocouple wires can result in significant heat loss from the thermocouple junction. To avoid the heat loss through conduction, sufficiently long and thin thermocouple wires are usually used. Heitor and Moreira suggested using a thermocouple length-to-diameter ratio of at least 200. (Heitor and Moreira, 1993) Another conduction consideration is that, for current experimental flame, in the central area, the temperature gradient along horizon direction was found to be very small, and the temperature profile was nearly one dimensional along vertical direction. In that case, error caused by conduction can be ignored when a long and thin thermocouple is used. Catalysis effects on thermocouple measurements in combustion environment are relatively ill-defined, both in terms of the magnitude of these effects and the conditions under which they are important. Catalysis-induced heating has most commonly been

84 68 ascribed to the effects of radical recombination on the surfaces of the thermocouple wire and bead (Heitor and Moreira, 1993). The principal means of eliminating the catalytic effect is to apply a non-catalytic surface coat onto the thermocouple. Silica (Kaskan, 1957), BeO/Y 2 O 3 ceramic (kent, 1970), and alumina-based ceramic (Burton et al., 1992) are reported to be used as non-catalytic surface coating materials. However, as a consequence, all these coatings would simultaneously affect the radiant, convective, and conductive heat transfer of the thermocouple. So it is really hard to perform a quantitative evaluation of the effects of catalysis itself. As for the temperature measurement experiment, there are two additional concerns about the coating: First, these coatings are fragile. To diminish soot accumulation effect, thermocouple was rapidly inserted into the flame. During this process, slight vibration was inevitable. Those fragile coatings could hardly stay long enough on the thermocouple to the end of measurement; second, these coatings can not withstand very high temperature. At 1900 K, stability of the coating became a problem. (Burton et al., 1992) During the experiment, before each temperature measurement at the desired location, soot particles accumulated on the thermocouple bead and wire were burned off by locating the thermocouple in the high temperature-oxidizing region at the edge of the flame, where the temperature was normally over 2000 K. Because of these concerns, uncoated bare thermocouples were used instead of coated ones.

85 69 Chapter 3 EXPERIMENTAL APPROACH 3.1 Materials During the initial investigation, we focused on the premixed ethylene/air flame, and ethanol and dimethyl ether were chosen as oxygenated fuel additives. Ethylene (C 2 H 4 ) was used as the fuel for the premixed flame study for two reasons. First, ethylene can produce enough soot in premixed flames, which makes it relatively easy to utilize optical diagnostic technologies; second, ethylene has been widely used in studies of soot formation for many years, so a large database exists upon which to build the mechanism of soot reduction would be more convenient. The structures of ethanol and DME are shown in Figure 3-1. Ethanol (C 2 H 5 OH) was selected as the primary oxygenated additive because it has been the subject of numerous combustion studies and has been investigated in diesel engines as an additive to reduce soot. The initial investigations into the use of ethanol in diesel engines were carried out in South Africa in the 1970s 127 and continued in Germany and the United States during the 1980s. 128 Most of these works showed a reduction in the smoke and particle levels emitted in the exhaust. This point, of increasing importance today, alone justifies the incorporation of ethanol into fuels.

86 70 H H O C H H C H H H H C H H O C H H Ethanol DME Figure 3-1 Chemical structure of ethanol and DME. 129 Dimethyl ether (CH 3 OCH 3 ) was also studied in the premixed flame to understand possible effects of the chemical structure of additive on soot reduction. Dimethyl ether (DME) has the same molecular formula as ethanol, but contains no carbon-carbon bond; the absence of C-C bond is believed to be responsible for the extremely low soot emissions from engines fueled with DME. 130 Under atmospheric pressure and temperature, DME is a gas with low toxicity and is relatively easy to handle. Also DME is a common substance used as an aerosol propellant (Hansen J.B 1995). Over the last ten years researchers have considered the use of DME as a fuel. It was found that DME did indeed effect a decrease in the emission of CO, NO x, formaldehyde, particulates, and nonmethane hydrocarbons, 131 compared with commercial diesel fuels. DME has also been successfully used as a methanol ignition improver in diesel engines where it has been reported to dramatically reduce total hydrocarbon emission. 132 A final consideration in the selection of ethylene, ethanol, and DME is the availability of detailed chemical reaction mechanisms that could be used in the explanation for their chemical effect on soot reduction.

87 Experimental Approach Figure 3-2 shows the schematic diagram of experimental setup. An unconfined flat laminar premixed flame was established using a McKenna Burner, which consists of a 60.2-mm-diameter water-cooled porous-plate. The flame was stabilized by a 100-mmdiameter, 10-mm-thick aluminum plate located 36 mm above the burner surface. The burner could be moved in vertical direction with a precision of 100 µm, using an adjustable jack. Figure 3-3 shows the schematic diagram of fuel, additive, and air supply systems. The purities of experimental gas and liquid are listed in Table 3-1 and they were used without further purification. The air and nitrogen to the burner were metered with rotameters. And the flow rates of ethylene and DME, which is gas phase at room temperature and pressure, were controlled by mass flow meters. All the rotameters and mass flow meters were calibrated with bubble meters. Liquid ethanol, which was supplied by a precision syringe pump, was vaporized and mixed with the ethylene in a vaporizer. The temperature of the vaporizer was maintained at 80ºC by electric heating tape, which is precisely controlled by an Omega temperature controller.

88 72 Burner: Premixed McKenna Burner Fuel: Ethylene (C2H4) Additive: Ethanol (C2H5OH) and DME (CH3OCH3) LII, LIF and LE Spectrometer PMT BOXCAR 532 nm for LII 266 nm for LIF Figure 3-2 Schematic diagram of experimental setup Preheat Fuel Cooling Water Inlet 53 o C Cooling Water Outlet 0.91 l/min Ethanol 80 o C Mass Flow Rotameter C2H4 Dry Air N2 DME Figure 3-3 Schematic diagram of fuel, additive, and air supply systems

89 73 Two mixing processes occurred in turn: first, ethylene mixed with additive in the vaporizer; then, the ethylene/additive mixture mixed with preheated air in a 160-cm-long mixing chamber. The mixing chamber consists of five equally spaced orifices mounted inside the chamber, and the length-to-diameter ratio of the chamber was roughly 80 in order to attain fully mixed conditions. To prevent any condensation of ethanol, the mixing chamber and fuel tube were heated with 5 heating tapes. By varying the electric voltage applied to the heating tape, the temperature of the flow at the lip of the burner was maintained at 53ºC for all mixtures regardless of whether a liquid additive was introduced. This approach ensures that all the flames were established at the same ambient condition. Mass flux of premixed ethylene/air flame (mg/s.cm 2 ) Equivalence ratio W/O Additive air ethylene (UHP) DME/ethanol % O in fuel air ethylene (UHP) DME/ethanol % O in fuel air ethylene (UHP) Table 3-1 Test Conditions in order to remain a stable flame, the mass flow rate at Ф = 2.64 is reduced to 80% of that at Ф=2.34

90 74 Test conditions in the current study are shown in Table 3-1 for a baseline fuel of ethylene with additive addition. Equivalence ratios studied were 2.34, and When an additive was introduced into the fuel line, the ethylene flow rate was reduced appropriately in order to keep the total carbon flow rate and equivalence ratio constant. Two oxygen contents, i.e., 5% and 10%, of ethylene/additive mixture, were fed to the fuel stream for each additive. For example, in the case of 5% oxygen addition, a mole fraction of of DME or ethanol in the fuel steam was required, which corresponds to a 14.2 % mass fraction of DME or ethanol in the fuel stream. Laser-induced Incandescence (LII) was applied to obtain spatial distribution of soot volume fractions in the flames. A doubled Nd:YAG laser (Surelite, Continuum) of 532 nm operating at 10 Hz was used as an intense light source to irradiate the soot particles. The induced incandescence at 90 o to the incident beam was imaged with a 105-mm UV camera lens (Nikon,f/4.5) onto an intensified charge coupled device (ICCD) camera (Princeton Instruments, Model ICCD-576S/RB). The camera gate time was set as 80 ns to include the laser pulse duration. A narrow band interference filter with the wavelength region of nm was placed in front of the camera to prevent laser scattering by soot particles from reaching the detector and to reject most background luminosity and laser-induced fluorescence. It is well known that LII signals can be categorized into two distinct regimes depending on the laser energy fluence, i.e., linear regime with respect to the laser energy fluence and a saturated regime regardless of the laser energy fluence. 91 In present study, the laser pulse energy fluence was measured approximately to be 0.6 J/cm 2, which is in the saturated LII regime. LII images were

91 75 acquired by averaging over 300 laser pulses. The uncertainties in the soot volume fractions derived from the LII measurements were less than 5%. A calibration factor from laser extinction measurement was used for the LII signals to obtain absolute local soot volume fraction. The extinction measurements were carried out using an Argon ion laser and a chopper (1KHz) lock-in amplifier. Since the flame was quite uniform in the radial direction and edge effects were negligible, data collection was simplified by taking line-of-sight average absorption measurements through the center of the flame. The nm laser beam was focused onto the flame using a 1-m focal length lens and received by a Silicon photodiode detector. Since laser beam kept fluctuating on the detector surface due to the large temperature gradient across the flame, an additional shorter focal length lens was employed to minimize the laser beam variation over the area of a photodiode detector. The system was calibrated against a series of Standard Natural Density Filters and the linearity was better than 99%. Data were recorded using a NI data acquisition system (NI-PCI-6110) with a sampling rate of 1000 samples/sec, and the mean voltage value was obtained by averaging over a period of one minute. In order to gain further insight into the chemical processes leading to soot reduction, measurements of PAH fluorescence were conducted. PAH species are often viewed as precursors to soot particles and possible sources of mass for soot growth. 133, 134

92 76 Figure 3-4 The size of PAH and their spectral properties. 135 In combustion environments, PAH often appear during the pyrolysis of hydrocarbons. PAH have a strong propensity to absorb ultraviolet (UV) radiation. Generally, the absorption and emission shift to longer wavelength, as the size of the aromatic structure increases, see Figure This property can be used to identify different PAH. 136 The association of similar broadband LIF with PAH species, observed in both premixed and diffusion flames, has been reported in the past (Coe, D.S., 1981; Miller, J.H., 1982; Prado, G., 1985; Beretta, F., 1985; Petarca, L., 1989). Some controversy remains regarding the correspondence between this broadband fluorescence and PAH species, since many possible classes of molecules can absorb and emit light in

93 77 the visible and near-uv regions, including single-ring aromatic, polycyclic aromatics, polyenes, and polyynes. All of these species have been detected in flame studies (Crittenden, B.D., 1973; Smyth, K.C., 1985; Hamins, A., 1990; D Alessio, A., 1992; Ciajilo, A., 1994). Evidence from combined optical and sampling measurements indicates that PAH are the dominant contributors to broadband fluorescence excited at either visible or near-uv wavelengths. Prado et al. 137 found a good correlation between the fluorescence intensity and PAH concentration profiles, while Beretta et al. 138 observed that their PAH fraction gave much stronger fluorescence than other classes of molecules sampled from premixed and diffusion flames. On the basis of those studies, the laserinduced broadband fluorescence in present study will be described as PAH fluorescence. The optical setup for PAH fluorescence measurement was very similar to that used for LII measurements, except that the fourth harmonic wavelength of Nd:YAG at 266 nm was employed in PAH LIF measurement. The UV beam of 266 nm excites both large and small PAH molecules. 139 By placing a combination of a 320 nm cutoff filter (WG320) and a UV pass filter (UG11) in the front of the camera, only the emission ranging from 320 to 380 nm, which corresponds to small PAH molecules (2 rings), was captured. In the same way, LIF emission of large PAH molecules (3 rings or more), ranging from 420 to 480 nm, was monitored by the camera after it passed through a combination of high pass filter (GG420) and low pass filter (BG12). The camera was gated-on for 35 ns during the laser pulse. Fluorescence from 500 laser pulses was accumulated to generate a PAH LIF intensity profile.

94 78 There are several difficulties in obtaining quantitative information from the LIF technique, including quenching, broadening of the absorption line, absorption of incident light and the self-absorption of fluorescence. 140 These effects tend to increase with pressure, and careful consideration is necessary to determine these parameters. 141 At this time there are no feasible method of calibration has been determined. Therefore, only relative concentration results of PAH were obtained. The spectra of both LII 142 and PAH 143 fluorescence are broadband in nature, and it would be difficult to isolate the fluorescence signal from LII emission by using filters alone. Fortunately, the laser excitation energy threshold of LII is much higher than that of LIF. For the PAH measurement, the energy of the pulse laser is 0.2 mj with duration of 10 ns and laser fluence of 1.0 x 10 5 W/ cm 2. Therefore, the excitation laser energy was kept well below the LII threshold, 2.5 x 10 7 W/cm Furthermore, the fluorescence signal was also detected in the presooting region (< 5mm), where the absorptivity and emissivity in the visible of PAH are very low, therefore it could not be incandescence. To verify the nature of the visible laser induced signal, under a similar condition, Sgro et al. 144 measured the time response of this signal at 530 nm, a parameter used to distinguish LIF from LII (Bengtsson and Alden, 1989; Vander Wal et al., 1997), at a height of 10 mm above the burner, well into the sooting region of the flame. The signal decay rate was an order of magnitude faster than that expected for LII signals, providing further evidence that the measured light comes from fluorescence rather than incandescence. The broadband spectral characteristics of the PAH fluorescence were observed using a 1/4-meter spectrometer (GCA-McPherson EU-700) over the spectral region from

95 to 600 nm, and corrected by spectrometer efficiency. On an optical axis perpendicular to the laser beam, a 100 mm focal length UV lens collected light over an 8 mm length of the beam in the flame and focused the image onto the entrance slit of the spectrometer. The calibration result using a 6035 Hg(Ar) lamp showed that the spectrometer had a resolution of 5 nm with a 0.4-mm-wide slit, which was positioned parallel to the propagation direction of the laser beam. And, the spectrometer was automatically driven by a Labview program at 5 nm/min. The signal from the photomultiplier (RCA Model 4840), which operated at 800 V, was processed by a BOXCAR and recorded using the NI data acquisition system with a sampling rate of 1000 samples/sec. The thermocouple probe configuration is shown in Figure 3-1. Sagging usually happens when a fine thermocouple is inserted into high temperature environment due to thermal expansion. The resulted position error could be 1 to 3 mm, depending on the temperature and the length of thermocouple legs. Cundy et al. 145 suggested a special configuration to avoid sagging of the thermocouple. The present study follows this configuration for thermocouple assembly.

96 80 Figure 3-5 Diagram of the Thermocouple Arrangement 146 The thermocouple used in this study is uncoated, pre-welded type R (Pt/Pt- 13%Rh) wire pairs, made by Omega. The junction was nearly spherical. Diameters of wire and junction bead are approximately 76µm and 160µm, respectively. Two hightemperature ceramic tubes were fixed with epoxy to each corresponding bolt head on the sliding mount. One bolt was held fixed by its nut, while the other was loose and allowed to pivot. A loose spring was then employed and stretched between the ends of the ceramic tubes approximately 40mm apart to provide tension on the thermocouple wire by pulling the ceramic tubes. This simple configuration gives advantage to avoid sagging of the thermocouple due to thermal expansion when the thermocouple is in the flame. The spring attached between tubes maintains constant tension and thus holds the accurate position of the thermocouple in the flame.

97 81 The junction temperature of thermocouple was obtained by rapid insertion of thermocouple to the flame by moving the thermocouple which was mounted on the plate back and forth on the rail. Before each temperature measurement at the desired location, soot particles accumulated on the thermocouple bead and wire were burned off by locating the thermocouple in the high temperature-oxidizing region at the edge of the flame. Temperature was measured up to the height of 20 mm above the burner surface. The starting point was 2 mm above the burner surface and the temperature measurement was taken with an interval spacing of 0.5 mm near the main reaction zone, where the temperature gradient is steep. In the post-flame zone where the temperatures were changing smoothly, temperature was measured with 1 mm or 2 mm intervals. The average insertion time to reach the final position was approximately 350 to 400 ms. Data points were recorded by NI data acquisition system. The final junction temperature was obtained from a linear extrapolation of the temperature history to time zero, representing the moment when the junction reached its final position. Due to the radiation heat loss from thermocouple, correction is needed to obtain accurate flame temperature. Assuming steady-state and negligible catalytic effect and conduction, the local gas temperature (T g ) can be obtained by the convective-radiative energy balance shown in Equation (3.1): T 4 = T T T d tc + ε tcσ ( 4 tc w (3.1) knu g )

98 82 where tc denotes the thermocouple junction. The correlation for Nusselt number used in the present study is: Nu = (0.4 Re d Re d ) Pr ( µ ) (3.2) µ s The gas properties such as thermal conductivity, specific heat and viscosity required for computing radiation correction were obtained using polynomial fit of individual gas constituent provided by CEA 1 program and database. 147 Then the gas properties were determined based on the estimated mole fractions of each gas constituent in the combustion product. The emissivity used in the present study is determined by the following polynomial fit: ε = *10 T 1.25*10 T *10 T (3.3) An extensive review and correlations for temperature correction procedures shown in Equations (3.1), (3.2), and (3.3) are presented in details in the work of Shaddix 148 and the present study followed the methodologies in his paper. Chemical Equilibrium with Applications, a program which calculates chemical equilibrium product concentrations from any set of reactants and determines thermodynamic properties for the product mixture

99 83 Chapter 4 EXPERIMENTAL RESULTS AND MODELING ANALYSIS FOR ETHANOL AND DME In this chapter, the experimental results of ethanol and DME addition in premixed ethylene/air flame are presented. First, measured flame temperature profiles are shown; then, PAH fluorescence results are presented, and followed by soot volume fraction results; after that, the effects of ethanol and DME on PAH and soot production are illustrated. A modeling study was also performed to further clarify the effects of oxygenated additives on PAH and soot production. The kinetic modeling work in the present study allowed identification of roles of oxygenated additives in the fuel during combustion by investigating chemical processes responsible for the formation and oxidation of soot precursors.

100 Experimental Results Temperature Profiles The measured flame temperature profiles are presented in Figure 4-1. The flame temperature was measured using thermocouples, and the results were radiation corrected following the method mentioned in former section. Two equivalence ratios were considered: Ф = 2.34 and At each equivalence ratio, 5% and 10% oxygen were introduced into the fuel by adding DME or ethanol. Temperature profiles are generally higher at Φ =2.34 than those at Φ =2.64. Also, at each equivalence ratio, the introduction of DME and ethanol resulted in higher flame temperatures than the corresponding baseline case.

101 Temperature (K) Baseline (Ф=2.34) DME 05% oxygen (Ф=2.34) ETOH 05% oxygen (Ф=2.34) DME 10% oxygen (Ф=2.34) ETOH 10% oxygen (Ф=2.34) Baseline (Ф=2.64) DME 05% oxygen (Ф=2.64) ETOH 05% oxygen (Ф=2.64) DME 10% oxygen (Ф=2.64) ETOH 10% oxygen (Ф=2.64) Height Above Burner Surface (mm) Figure 4-1 Temperature profiles of premixed ethylene/air flame, with and without ethanol and DME addition

102 86 Generally, the higher the soot concentration, the more energy is radiated by soot, and results in lower flame temperature. Based on this point of view, the increase of temperature profiles by introducing ethanol and DME is consistent with the soot volume fraction measurement results shown later in Figure In Figure 4-18, with the introduction of 5% oxygen into the fuel by the addition of ethanol at Ф=2.34, a reduction of 25% in soot volume fraction is observed; and as for DME, about 10% more soot reduction is observed. As a result, DME shows the highest temperature profile due the lowest soot concentration. However, the increase of peak temperature due to the introduction of oxygenates seems to be contradictory to the peak temperature prediction provided by CEA program and database, 147 which showed a reduction of peak temperature by introducing both DME and ethanol under constant equivalence ratio. It can be seen that, in Figure 4-1, the peak temperature points are within the height range of 4 to 6 mm above the burner surface, where the soot starts to appear and the soot volume fraction is relatively low, as shown later in Figure Within this height range, the radiation loss by soot is not sufficient to explain the contradictory trends of temperature profiles. This discrepancy may be caused by the different initial temperatures of the ethylene/air mixture before the reaction starts. The adiabatic flame temperature predicted by CEA is a function of the initial temperature. According to the CEA prediction for present ethylene/air flame, an increase of initial temperature from 300 K to 400 K can result in an increase of 70 K for the predicted peak temperature. However, in present prediction, a constant initial temperature, 300K, was used. This may not be true. During

103 87 the experiments, the initial temperature of the ethylene/air mixture is primarily controlled by the temperature of burner surface through heat conduction; while the stabilizing plate, sitting 36 mm above the burner surface, is a sound source for providing heat to the burner surface by radiation, since the plate is covered by black soot. After the addition of additive, the increase of flame temperature near the plate will increase the temperature of the plate. Therefore, more energy is radiated from the stabilizing plate to the burner surface, and increases the burner surface temperature. Consequently, the initial temperature of premixed ethylene/air is increased by heat conduction when the mixture passes through the burner surface. Those observations are also consistent with the temperature measurement results of Inal and Senkan. 149 Recently, Inal and Senkan investigated the effects of three oxygenated additives, methanol, ethanol and MTBE, on the formation of PAH and soot in laminar, premixed, atmospheric fuel rich n-heptane (n-c 7 H 16 )/Ar/O 2 flames. The equivalence ratio selected in their study was 2.10 and the oxygen weight in the fuel was kept at 2.7 percent for each n-heptane/oxygenate mixture. Same thermocouple was used in their experiment as in the present study, except that their thermocouple was silicon oxide-coated. To minimize the thermocouple exposure to soot, the same rapid insertion technique was used. Their temperature profiles for the n-heptane and n- heptane/oxygenate flames are presented in Figure 4-2, and those are direct thermocouple readings without radiation correction. As can be seen from this figure, the peak temperatures of all the oxygenate-containing flames were higher than the n-heptane flame. The peak temperature was 1528 K for the n-heptane flame, while the peak temperatures for the flames containing methanol, MTBE, and ethanol were 1550, 1565,

104 88 and 1571 K, respectively. However, according to the prediction of CEA, with a constant initial temperature of 300 K, the addition of ethanol and methanol would drop the peak temperature for about 18 and 14 K, respectively, compared with the baseline n-heptane flame. These consistencies provide further support for present temperature measurements. Figure 4-2 Flame temperature profiles from Intal and Senkan. 149

105 89 Figure 4-3 presents the digital pictures of flames corresponding to each condition in Figure 4-2. The luminosity of the flame reflects the radiation from soot particles, and the more soot in the flame, the brighter the flame. In Figure 4-2, it can be seen that the flames at Ф=2.64 are generally brighter than those at Ф=2.34, indicating more soot at higher equivalence ratio. Also, at each equivalence ratio, the luminosity of flame drops after the addition of additives, which means additives did affect those flames, and reduced the soot load, though it is hard to identify quantitatively from the luminosity of these pictures. It can also be seen that there is a gap between burner surface and the luminous region of the flame, which indicates that there is no soot at the location near burner surface, and the soot begins to appear after that. In Figure 4-3, it can be found that the luminous region begins at different locations for the baseline flame and the additive-containing flames for Ф=2.64. The more additive is added, the bigger the gap. Meanwhile, in Figure 4-1, the measured temperature profiles for Ф=2.64 show a substantial drop with the addition of additives, for the height range of 2 to 4 mm above the burner surface. This coincidence may be explained by the flame speed: the higher the flame speed, the closer the flame sitting on the burner surface. The addition of additives at Ф=2.64 may reduce the flame speed, and as a result, the flame shifts up. Therefore, lower flame speed results in a bigger gap between the burner surface and the luminous region of the flame. At the same time, the temperature profiles also shift up toward higher position, and result in a large drop of temperature profiles due to this flame shift.

106 Figure 4-3 Pictures of the ethylene/air flames with and without additives. (Digital Camera Setting: F=2.8, Time=1/800 s) 90