Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Particulate matter measurement in a shock tube facility under engine-relevant conditions Wang, Timothy Xi 2007

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


831-ubc_2007-0270.pdf [ 17.69MB ]
JSON: 831-1.0080737.json
JSON-LD: 831-1.0080737-ld.json
RDF/XML (Pretty): 831-1.0080737-rdf.xml
RDF/JSON: 831-1.0080737-rdf.json
Turtle: 831-1.0080737-turtle.txt
N-Triples: 831-1.0080737-rdf-ntriples.txt
Original Record: 831-1.0080737-source.json
Full Text

Full Text

P A R T I C U L A T E M A T T E R M E A S U R E M E N T IN A S H O C K T U B E F A C I L I T Y U N D E R E N G I N E - R E L E V A N T CONDITIONS by TIMOTHY XI WANG B.A.Sc., The University of Waterloo, 2001 A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF MASTER OF APPLIED SCIENCE in T H E F A C U L T Y OF G R A D U A T E STUDIES (MECHANICAL ENGINEERING) THE UNIVERSITY OF BRITISH COLUMBIA April 2007 © T i m o t h y Xi Wang, 2007 Abstract This study develops and demonstrates a particulate matter (PM) measurement system in a shock tube facility, in order to investigate correlations between P M emissions and combustion parameters. The resultant method was applied to premixed and non-premixed experiments using several diesel-alternative gaseous fuel mixtures. As the main component of shock tube P M , soot formation mechanisms are highly complex. The sampling methodology evolution considered technical challenges in experimental conditions attainment, contamination control, particle loss minimization, and proper instrument detection. The key resultant sampling system apparatus and procedures include conductive surfaces, particle impaction, and tube settling. Consistent back-ground black carbon (BC) levels of 100-150 ng have been achieved in blank tests. Significant particle losses through visible B C mass increment curve decay have also been eliminated for both blank and injection experiments. Aethalometer data analysis algorithms are modified to suit the needs and limitations of this novel experimental setup. The preliminary results under engine-relevant conditions show the promise of methane/natural gas in meeting the 2007 standards. The limited set of premixed experi-ments (using methane and methane/ethane) did not produce noticeable B C trends with combustion temperature, pressure, or equivalence ratio (EQR) . Non-premixed experi-ments with a gaseous fuel (using methane/DME, methane, and methane/ethane blends) injector also lacked clear dependences on temperature or pressure. Larger experimental sets of low E Q R premixed and higher fuel mass injections should produce meaningful results. Dominant errors are due to particle loss and optical specific attenuation uncer-tainties. External measurement validation and shot-to-shot variability must be studied ; for proper B C signal interpretation. It is extremely challenging to achieve accurate and re-peatable global B C mass measurements from methane flames in a shock tube. Any future work to build upon the current sampling system and methodology should be carefully approached. i i i Table of Contents Abstract . ii Table of Contents . ' . . i i i List of Tables vii List of Figures . ix List of Symbols and Abbreviations xi i Acknowledgements xv 1 In t roduc t i on 1 2 L i t e ra tu re R e v i e w 3 2.1 Introduction 3 2.2 Background 4 2.3 Soot Formation Mechanisms 5 2.4 Direct-Injection Natural Gas Engines . . . . 8 2.5 Particulate Matter Sampling from Shock Tubes 9 2.5.1 Shock Tube Dynamics 9 2.5.2 Particulate Matter Sampling Instrumentation 12 2.5.3" In-Situ P M Experiments in the Shock Tube . 14 2.5.4 Shock Tube Experiments of P M Emissions 15 2.6 Conclusions 17 3 Deve lopment of E x p e r i m e n t a l M e t h o d o l o g y 19 3.1 Introduction 19 3.2 Description of Facility 20 3.3 Achieving Desired Experimental Conditions 23 3.3.1 Elimination of Pressure Disturbances 24 3.4 Contamination Control . . . 25 iii Table of Contents j v 3.4.1 Contamination Detection • 26 3.4.2 Pre-Experiment Control " 28 3.4.3 Post-Experiment Control 34 3.4.4 Particle Identification 39 3.5 Particle Loss Minimization . . . 45 3.5.1 Inside Shock Tube . 46 3.5.2 Gas Venting Process 49 3.5.3 Sample Container 50 3.5.4 Container to Aethalometer 52 3.6 Instrument Detection 54 3.6.1 Operation Principle 55 3.6.2 Data Algorithm Modifications 57 3.6.3 Error and Uncertainty 61 3.7 Conclusions • 64 4 Particulate Matter Sampling from Methane Flames 67 4.1 Introduction 67. 4.2 Premixed Study . . . 68 4.2.1 Apparatus 68 4.2.2 Procedure 70 4.2.3 Results and Discussion -. 73 4.2.4 Error Analysis 78 4.3 Non-Premixed Study 80 4.3.1 Apparatus . 81 4.3.2 Procedure . 83 4.3.3 Results and Discussion 84 4.3.4 Error Analysis ; 92 4.4 Conclusions ' 93 5 Conclusions and Recommendations 96 References 99 Appendices 104 Table of Contents 104 104 104 105 107 109 111 111 B Diaphragm Material Selection and Testing 115 B. l Summary . . 115 B.2 Material Selection 115 B.3 Burst Pressure Testing 116 B. 4 Determination of Throttl ing Losses 117 C Contamination Control 121 C. l ' Contamination Detection 121 C.2 Pre-Experiment Control 122 C.3 Post-Experiment Control 122 C. 4 Particle Identification : 127 C.4.1 S E M Analysis 127 C.4.2 T E M Analysis .• - 130 D Particle Loss Minimization s 138 D. l Inside Shock Tube 138 D.2 Gas Venting Process 140 D. 3 Sample Container 140 E Aethalometer Data Analysis 141 E. l Operation Principle . . 141 E.2 Data Algorithm Modifications . . . 144 E. 3 Instrument Maintenance 146 F Aethalometer Flow Meter Correction and Calibration 155 F. l Flow Rate Correction 155 A Background A . l Soot Formation Mechanisms A. 1.1 Soot Precursors . . . A. 1.2 Particle Inception . . A. 1.3 Particle Growth . . . A. 1.4 Oxidation A.2 Shock Tube Apparatus . . . A.3 Previous Work Table of Contents ' v j F . l . l Direct Correction 156 F. 1.2 Manufacturer-Suggested Correction 157 F. 2 Experimental Calibration 158 G P r e m i x e d E x p e r i m e n t D a t a 165 G. l Procedure 165 G.2 Results 165 G. 3 Error Analysis 166 G. 3.1 Driven Gas Composition 167 G.3.2 Experimental Temperature/Pressure . . . 170 G.3.3 Black Carbon Mass 170 H N o n - P r e m i x e d E x p e r i m e n t D a t a 177 H. l Procedure 177 H.2 Results 177 H.3 Error Analysis 177 List of Tables 3.1 Contamination Detection 27 3.2 Summary of Pre-Experiment Control Procedures 34 3.3 Summary of Post-Experiment Control Procedures 39 4.1 Comparison of Standard and Measurement Errors 80 A. l Summary of In-Situ and P M Emissions Experiments 112 B. l Summary of Burst Pressure Testing 120 B. 2 Summary of Diaphragm Polytropic Ratios 120 C. l Summary of Total Particle Counts 122 C.2 Elemental Composition of Various Materials 130 C.3 S E M Analysis of Lacy Carbon Gr id from Blank Experiment . 136 C.4 S E M Analysis of Lacy Carbon Grid from Sooty Experiment 136 • C.5 T E M Analysis of Lacy Carbon Gr id from Blank Experiment 137 C. 6 T E M Analysis of Lacy Carbon Gr id from Sooty Experiment 137 D. l Summary of Gas Venting Durations 140 E. l Summary of Specific Attenuation Values 142 E.2 Sample Black Carbon Channel Data 148 E.3 Attenuation Value Changes Due to Gas Composition 149 E.4 Sample Post-Processed Aethalometer Data 150 E.5 Exponential Decay Test Data 151 E.6 Sample Post-Processed B C and U V Channel Data . : 153 E. 7 Sooty Experiment B C Data 154 F. l Calibrated Flow Rates (LPM) . . , 159 vii List of Tables vii i F. 2 Calibration Data . . . . 160 G. l Premixed Methane Series Experimental Test Matr ix 165 G.2 Premixed Methane/Ethane Series Experimental Test Matr ix 168 G.3 Premixed Blank Series Experimental Test Matrix 168 G.4 Premixed Methane Series Conditions . 169 G.5 Premixed Methane/Ethane Series Conditions 172 G.6 Premixed Blank Series Conditions 173 G.7 Premixed Methane Series Aethalometer Conditions 174 G.8 Premixed Methane/Ethane Series Aethalometer Conditions 175 G. 9 Premixed Blank Series Aethalometer Conditions 176 H. l Non-Premixed Methane/DME Series Experimental Test Matr ix 179 H.2 Non-Premixed Methane Series Experimental Test Matr ix 179 H.3 Non-Premixed Methane/Ethane Series Experimental Test Matr ix 180 H.4 Non-Premixed Methane/DME Series Conditions 181 H.5 Non-Premixed Methane Series Conditions 182 H.6 Non-Premixed Methane Series Conditions (Continued) 183 H.7 Non-Premixed Methane/Ethane Series Conditions '184 H.8 Non-Premixed Methane/DME Series Aethalometer Conditions 185 H.9 Non-Premixed Methane Series Aethalometer Conditions 186 H.10 Non-Premixed Methane Series Aethalometer Conditions (Continued) . . 187 H . l l Non-Premixed Methane/Ethane Series Aethalometer Conditions 188 List of Figures 2.1 U.S. Heavy-Duty Diesel Engine Emissions Regulations [53] . 5 2.2 Aggregates of Primary Soot Particles [13] 6 2.3 Summary of Soot Formation Stages [51] . . . 7 2.4 Shock Tube Principle [19] 11 3.1 Shock Tube Apparatus Schematic [19] . 22 3.2 Double Diaphragm Setup [21] ; 22 3.3 Shock Tube Facility 23 3.4 Light and Pressure Signals from Combustion Test (20 bar, 1200 K) . . . 29 3.5 Frames of Pre-Injection, Fuel Ignition, and Flame Propagation 30 3.6 Rubber Gasket Particle 41 3.7 Rubber Gasket Composition 42 3.8 Stainless Steel Diaphragm Particle 43 3.9 Stainless Steel Composition 44 3.10 Lexan Particle (1 ^m) 45 3.11 Soot Particle 46 3.12 Electrostatic Loss Comparison for Various Bag Materials 52 3.13 Conductive vs Non-Conductive Line Loss Experiment 53 3.14 Long vs Short Line Loss Experiment . . 54 3.15 Data Spikes Due to Gas Composition Changes 58 3.16 Typical Black Carbon Sampling Results . 61 3.17 Typical Black Carbon and U V P M Sampling Results 62 4.1 Premixed Particulate Matter Sampling Apparatus 70 4.2 Premixed Methane Series Results 74 4.3 Premixed Methane/Ethane Series Results 75 ix List of Figures ' x 4.4 Blank Test Results for Premixed Series 76 4.5 Non-Premixed Particulate Matter Sampling Apparatus 82 4.6 Non-Premixed Methane/DME Series Injection Results 85 4.7 Non-Premixed Methane/DME Series Blank Results . 86 4.8 Non-Premixed Methane Series Injection Results 87 4.9 Non-Premixed Methane Series Blank Results 88 4.10 Non-Premixed Methane/Ethane Series Injection Results 89 4.11 Non-Premixed Methane/Ethane Series Blank Results . 90 4.12 Comparison of Non-Premixed Blank Experiment Results 91 A . l Precursor Formation Process [51] 105 A.2 Particle Inception Process [15] 107 A.3 Quantitative Soot Parameters During Inception and Growth [15] 109 A.4 Injector Connection Setup [21] I l l A.5 Optical Access Section [21] . . . 113 A.6 Optical Access Principle [19] . . . 113 A. 7 Shock Tube Sampling at the University of Illinois at Chicago [5] 114 B. l Stainless Steel Shim Diaphragm 116 B.2 Carbon Steel Shim Diaphragm 116 B.3 Lexan Diaphragm . 117 B.4 Pressure Disturbances Within Experimental Region (40 bar) 118 B.5 Diaphragm Fragment Trapping Devices 119 B. 6 Throttl ing Losses [19] ' ' . . 119 C. l Passing Through Shock Tube , 121 C.2 Bypassing Shock Tube ' 122 C.3 Blank Test Light Emission (20 bar, 1200 K) . 123 C.4 Inert Blank Test Light Emission (20 bar, 1200 K ) 123 C.5 Comparison of Tube Surfaces 124 C.6 Towel and Brush Cleaning . . . . 124 C.7 High-Speed Rotating Brush Cleaning Setup [41] 124 C.8 Gas Jet Cleaning Tool 125 C.9 Magnetic Filter 125 C.10 Schematic of Hypersonic Impactor 126 List of Figures xi C . l l Impactor Location 127 C.12 Multi-Stage Impactor Schematic ' . . .. 128 C.13 Single Stage Multi-Nozzle Impactor: Nozzle and Impaction Plates . . . . 128 C.14 Dust Particle 129 C.15 Dust Composition 129 C.16 T E M Sampling Setup . . 130 C. 17 Background Laces 131 C.18 Lexan Particle (100 nm). • • 131 C.19 Lexan Particle (200 nm) 132 C.20 Lexan Particle (150 nm) 132 C.21 Lexan Composition • • 133 C.22 Soot Particle 2 : 133 C.23 Soot Particle 3 134 C.24 Soot Composition 134 C.25 T E M Soot Photograph 1 135 C.26 T E M Soot Photograph 2 . . 135 E . l Exponential Decay Test Curve Fit . 146 E.2 Linear Curve Fi t 147 E.3 Polynomial Curve Fi t (Type 1) 147 E.4 Polynomial Curve Fit (Type 2) ' . . . 149 E.5 Steady Mass Increment Curve 152 E. 6 Sooty Experiment B C Mass Increment Curve 152 F. l Flow Meter Calibration Apparatus 158 F.2 Linear Flow Rate Correlations 163 F. 3 2-D Surface Interpolation 164 G. l Dynamic Pressure Transducer Signals 166 G. 2 Incident Shock Velocity Calculation . 167 H. l J-43 Mass Flow Correlation 178 List of Symbols and Abbreviations 7T pi A Mean Free Path rj Gas Viscosity a Specific Attenuation Coefficient p Density 7 Specific Heat Ratio a Local Speed of Sound amu . Atomic Mass Unit A Hamaker Constant A T N Optical Attenuation Unit bhp-hr Brake Horsepower-Hour B C Black Carbon cm Centimeters C Carbon Cp Specific Heat at Constant Pressure . C C D Charge-Coupled Device C M O S Complementary Metal Oxide Semiconductor C N G Compressed Natural Gas C P C Condensation Particle Counter d Particle Diameter D Diffusion Coefficient D M A Differential Mobil ity Analyzer D M E Dimethyl Ether D M P S Differential Mobility Particle Sizer E D X Energy Dispersive X-ray E P A Environmental Protection Agency xi i List of Symbols and Abbreviations xiii EQR Equivalence Ratio Soot Volume Fraction Fadh Adhesive Force Thermophoeretic Force g Grams H Hydrogen HPDI High Pressure Direct Injection ID Internal Diameter k Specific Heat Ratio L Liters L P M Liters per Minute m Meters mg Milligrams mm Millimeters ms Milliseconds mL Milliliters Ma Mach Number ng Nanogram nm Nanometer N Particle Number Density NG Natural Gas NO, Oxides of Nitrogen 0 Oxygen psi Pounds per Square Inch P Pressure PAH Polycyclic Aromatic Compounds P M Particulate Matter P M 1 0 Particulate Matter Under 10 (im Diameter Q Flow Rate rpm Revolutions per Minute R Universal Gas Constant RMS Root Mean Squared s Seconds List of Symbols and Abbreviations SEM Scanning Electron Microscopy SOF Soluble Organic Fraction SS Stainless Steel T Temperature TEM Transmission Electron Microscopy TEOM Tapered Element Oscillating Microbalance TOR Thermal Optical Reflectance Microgram fim Micrometer UBC University of British Columbia UV Ultra-Violet V Volts Vth Thermophoeretic Velocity . Terminal Settling Velocity X Separation Distance y Mole Fraction J Acknowledgements I would like to thank, first of all, my supervisors Dr. Steve Rogak and Dr. Kendal Bushe, for the opportunity to conduct world-class research under their guidance and expertise.. Their dedication and support throughout this project have contributed greatly to its accomplishment, and I am grateful for their patience, understanding, professional advice, and funding. I also wish to thank Dr. Gregory Sullivan for his technical knowledge and contribution on the experimental challenges encountered. In addition, I am indebted to the UBC mechanical engineering department for providing the state-of-the-art facilities and functions, as well as to the respective faculty, staff, and technicians for their invaluable assistance throughout my coursework and research program. Much ap-preciation also goes to Westport Innovations for their industrial and financial partnership. I am also fortunate to have met many talented and hardworking students during my studies here. It has been an enjoyable experience to work with my colleagues Jian Huang, Jean-Louis Iaconis, Heather Jones, Mohammad Khan, Davood Faraji, Bulent Guzel, Anthony Huen, Olukayode Jinadu, and Edward Chan. I would like to especially thank Jian and Jean-Louis for their assistance on the shock tube theory and operation, Heather for the P M collection and microscopy advice, and Mohammad and Davood for their continual experimentation support. The mutual friendships and encouragements have helped me complete this challenging but rewarding journey. Last but not least, I thank my friends and family for their continual acceptance of the sacrifices in achieving my academic goals. The tremendous support and patience from everyone during this time have provided me extra strength and motivation, for which I am truly grateful. Special mention goes to Masaki for accompanying me through the tough times and for reminding me of the true perspectives in life. I would like to, in the end, dedicate this accomplishment to my parents, whom I couldn't have done this without. My sincere appreciation to them for enduring the countless sacrifices over the years, for understanding my motivation, ambition, and drive to undertake this step, for encouraging me to never settle for the path of least resistance, and most importantly to believe in my goals and abilities. I know they will be as proud as I am in achieving this small but significant destination in my life. xv Chapter 1 Introduction The motivation to develop cleaner-burning fuels has been accelerated by increasingly stringent governmental regulations. As conventional diesel engines are notorious polluters of particulate matter (PM) and oxides of nitrogen ( N O x ) , vast amounts of research are being focused on relative emissions characteristics of potential alternative fuels, without compromising engine performance. For example, Westport Innovations' High Pressure Direct Injection (HPDI) technology utilizes natural gas as the main fuel to reduce P M and N O x emissions while maintaining the inherent diesel engine benefits of torque, power, and efficiency. The ultimate goal of such research is therefore to find the optimum engine operating parameters for the given fuel to minimize P M emissions. However, due to the presence of lubricating oil and pilot fuel, as well as variables such as engine geometry and unsteady cylinder conditions, direct P M sampling from engines presents significant challenges in differentiating the fundamental underlying relations between soot emissions and combustion conditions. A simplified combustion apparatus such as a shock tube can reduce external variable effects and interactions on the measured quantities. This facility allows other variables to be kept relatively constant while investigating cause-effect relationships. The purpose of this study is to explore possible correlations between soot emissions and combustion parameters in a shock tube. Methane-based combustion under diesel engine-relevant conditions will be used as a preliminary study of the P M measurement methodology. It is hoped that the results can be used to complement actual engine testing and numerical predictions to eventually determine the optimum engine operating conditions for P M minimization, as well as to provide insight into engine and injector designs. 1 1. Introduction 2 The extraction and measurement of combustion emissions (especially P M ) from shock tubes are novel techniques. In-the absence of directly applicable experimental methodol-ogy, literature review is limited to the relevant, shock tube exhaust gas sampling, general particle sampling, and sampling instrumentation considerations. Background information on soot and its formation process is also provided to emphasis the complex chemical and physical mechanisms, as well to help explain the preliminary results from methane flames. The next phase of the study involves the development of an appropriate P M sampling system and methodology. The U B C shock tube facility and its associated experimental challenges are first described. Key technical challenges based on environment control, particle dynamics, and experimental hardware, are carefully analyzed. In addition, detailed instrumentation considerations contribute significantly to the development process. Representative blank, premixed, and injection experiments are used throughout this phase to identify possible additional challenges, as well as to validate the effective-ness of any system modifications. The purpose of this methodology development is to critically evaluate and analyze each and every possible sampling path, and to devise solutions to ensure that the measured P M quantities are representative of those from the actual combustion event. Finally, preliminary P M measurements from methane flames are performed, under both premixed and non-premixed environments. Premixed experiments are used primarily to generate sufficient soot to aid the methodology development, and at the same time to probe possible trends of P M versus combustion temperature, pressure, equivalence ra-tio, and fuel mass. Non-premixed experiments, using a Westport J-43 injector, are used to investigate correlations of P M versus temperature, pressure, fuel mass, and injection duration. For each flame type and experiment series, the same parameter is varied while other variables are held as constant as possible. The purpose of this preliminary experi-mental study is to test the practicality of the P M measurement system, provide insight into system shortcomings/improvements, and to note any discernable trends (i.e. P M ver-sus combustion conditions) for detailed future experimental investigations in both shock tube and engine settings. Chapter 2 Literature Review 2.1 Introduction Particulate matter (PM) from hydrocarbon combustion processes is composed of two main components: soot and semi-volatile material. The semi-volatile component is made up of condensable organic material from products of incomplete combustion such as unburned lubricating oil and unburned fuel. Soot, however, refers to the PM component from the direct chemical and physical mechanism conversion of hydrocarbon fuel to solid molecular carbon. Although both components are physically detected as PM mass, the soluble organic fraction or SOF (percentage of semi-volatile mass to total PM mass) is difficult to determine precisely, due to complex interactions from their formation stages to the final measurement (e.g. a carbon core immersed in semi-volatile PM). More importantly, the behavior of these PM components is greatly affected by engine variables such as flow geometry, external ignition sources, multiple fuel jets, and unsteady cylinder temperatures and pressures. Since the shock tube provides an isolation facility to simulate engine combustion without the semi-volatile material sources, soot will likely be the only PM component in the combustion products. The additional capability of the shock tube to eliminate most external variables in the engine setting also allows potentially better attributions of soot output and combustion parameters. In order to meet future PM emissions standards, these fundamental correlations must be investigated to aid the more daunting tasks of minimizing actual engine emissions. Due to the nature of the shock tube PM, this literature review will initially focus on soot formation processes and mechanisms. Recent developments in natural gas-fueled engines 3 2.2. Background 4 operating on a diesel-cycle will be discussed in the context of Westport's High Pressure Direct Injection technology and their relevance to the shock tube. After an overview of general shock tube principles, conventional P M sampling methods will be surveyed for the purposes of the upcoming sampling system development. Previous work on shock tube P M experiments and sampling are subsequently summarized. As the principal fuel of this investigation, methane-based combustion experiments in shock tube settings will be described in detail. Since the reported soot masses from the most comparable literature experiments are very small, the suitability and operating principles of various sensitive sampling instrumentation are reviewed with regards to the study objectives. 2.2 Background Soot formation from hydrocarbon combustion processes have been extensively studied experimentally and numerically. The detailed mechanisms of soot formation and oxida-tion are still not fully understood, partly due to the fact that it is much more sensitive to process conditions than other combustion processes. Although numerical simulations of soot production in the flame can be more readily tweaked for new fuel species, the often-simplified underlying chemical mechanisms render most predictive models inadequate. As a result, experimental measurements of soot (either in-situ or from exhaust dumping) are usually the only source of reliable data [29]. The experimental results can also be complementary to the theoretical considerations in predicting and correlating soot emissions in new engineering applications. In conventional diesel engines, soot is the main particulate matter component because fuel-rich regions always exist in diffusion flames. Although soot and P M will often be used interchangeably in this study, both P M species mentioned above contribute to the host of adverse health and environmental effects that led to the ever-increasing emissions standards. For example, combustion-related P M is one of the major sources of PMio (under 10 / im in diameter) known for being carcinogenic, as well as causing respiratory problems. In addition, along with N O x emissions, combustion P M contributes to smog and global warming. It is estimated that U.S. diesel powered vehicles produce 10 1 1 g of soot annually [44], Soot is also called black carbon (BC) due to the predominant carbon component and the observed emitted color. Since the only known sources of soot 2.3. Soot Formation Mechanisms 5 particles in the atmosphere is the combustion of carbonaceous fuels [14], regulatory bodies such as the U.S. Environmental Protection Agency (EPA) have significantly increased the stringency of the standards for engine emissions. For example, allowable heavy-duty diesel engine emissions for both P M and NOx have dropped an order of magnitude from 2003 to 2007, as shown in Figure 2.1. Figure 2.1: U.S. Heavy-Duty Diesel Engine Emissions Regulations [53] 2.3 Soot Formation Mechanisms Most of the available information on the fundamentals of soot formation in combustion comes from studies in simple premixed and diffusion flames, stirred reactors, shock tubes, and constant-volume combustion bombs. The formation of soot is one of the most complex chemical phenomena during hydrocarbon combustion, because the process is affected by a wide array of conditions and parameters such as pressure, temperature, fuel, oxidizer, radical levels, and geometry [42]. Essentially, low molecular weight (gaseous) hydrocarbons are converted to mostly solid carbon within extremely short time scales. The various homogeneous and heterogeneous processes involved are both kinetically and thcrmodynamically controlled, as shown in Appendix A. Quantitatively, soot volume fraction (v (ratio of soot volume to total volume), particle number density N (ratio of soot particle number to total volume), and the average particle diameter d are typically used to characterize the formation stages. 2.3. Soot Formation Mechanisms G The primary soot particles formed in combustion processes are very similar, with mean diameters of tens of nanometers. They contain at least 1% hydrogen by weight, and correspond roughly to an empirical formula of C 8 H [13]. Each primary particle is made up of a large number (104) of crystallites, and contains up to 106 carbon atoms. The actual chemical structure is a multiple ring polynuclear aromatic compound, and is difficult to characterize because of the unclear transition from gas to liquid to solid phases. These roughly spherical particles (or spherules) often attach upon formation, forming long branched or straight chain aggregates as shown in Figure 2.2. The smallest soot particles are formed in luminous but non-sooting flames, while the largest are obtained in heavily sooting flames (i.e. the soot escapes the flame). One theory for the similarities in various soot particle properties pertains to the increasing C / H ratio of the virgin soot particles as they pass through the hot burnt gases [13]. The fundamental processes governing the formation of soot particles also share a common sequence of events [43] summarized below and detailed in Appendix A. i) A chemical kinetically controlled reaction sequence which results in the formation of precursor species, ii) A particle inception stage which results in the formation of large numbers of small primary particles, iii) Particle growth stage in which surface growth and particle coagulation processes increases the particle size, and iv) A stage in which material is no longer added and the particle size is both increased by agglomeration and reduced by oxidation. Figure 2.2: Aggregates of Primary Soot Particles [13] Figure 2.3 provides a graphical summary of these formation stages. In diffusion flames 2.3. Soot Formation Mechanisms .7 (e.g. diesel engines, gas turbines, turbulent jet flames, etc.), soot forms in the locally fuel-rich regions, and generally yields higher volume fractions than premixed flames. Temper-ature has a major physical effect on soot formation. Increasing the temperature under premixed conditions generally increases soot production, whereas the opposite is true for diffusion flames [13]. The effects of pressure are felt through the changes in system tem-perature, species concentration, and the associated shifts in equilibrium. The final soot volume fraction increases with increasing pressure and increaing C / O ratio [51]. The crit-ical C / O ratio for soot formation also increases with increasing temperature but is weakly dependent on pressure. Beyond the carbon formation limit, soot yield increases rapidly with increasing C / O ratio and is strongly enhanced by pressure increases. The effect of other common parameters on soot production rates can be found in Appendix A. The rel-ative, importance of the various stages of soot formation is often system-dependent [29], which complicates general understandings of individual combustion parameter effects. The slowest (rate-controlling) process in the overall mechanism is likely the formation of the first aromatic ring structure, which explains why aromatic fuels have the greatest sooting tendency. This builds on the evidence that fuel pyrolysis rates and mechanisms control the overall soot formation process and therefore the tendency of a given fuel to soot. Most importantly, all soot particles created in combustion continuously experience a competition of growth and oxidation processes, thus complicating the understanding and relation of soot formation to eventual emission characteristics. Reaction time f d-50nm Coagulation Surface growth and coagulation Particle inception Particle zone —' ••— d-0.5nm Molecular zone Figure 2.3: Summary of Soot Formation Stages [51] 2.4. Direct-Injection Natural Gas Engines 8 2.4 Direct-Injection Natural Gas Engines Development of natural gas (NG) fueled internal combustion engines has been ac-celerated in recent years due to its attractive emissions reduction potential, low fuel cost, as well as being less reliant on decreasing crude oil resources. There have also been optimistic projections on the widespread use of natural gas based transportation engines and the associated fueling infrastructure, as part of the imminent alternative fuel revolution. However, in order to utilize the torque, power, and thermal efficiency benefits of conventional compression-ignition (late-cycle injection) engines, inherent differences between the fuels must be addressed. The main drawback of directly injecting natural gas is the prohibitively long ignition delay time of methane. Current approaches to minimize the ignition delay of injected natural gas include the use of dual fuels, ignition-promoting additives, glow/spark plugs, and increased compression ratios to achieve higher temperatures [21]. The High Pressure Direct Injection (HPDI) technology developed by Westport Innova-tions attempts to combine some of these approaches to achieve operational feasibility of N G injection, while realizing the low-PM producing nature of gaseous fuels. The HPDI system utilizes a small amount of diesel-fuel ( ~ 2-5% of the total fuel energy) ignite the natural gas charge. The diesel pilot, with a much shorter ignition delay, is used solely for igniting the main fuel immediately after N G injection, analogous to a spark-ignition setup. However, the presence of the diesel, along with the engine's lubricating oil, signif-icantly complicates the attribution and interpretation of the sampled P M emissions [25]. As diesel is inherently a higher-PM producing fuel, its P M component overshadows the total mass sampled, and masks the extent of achievable P M reduction due to N G use. The lubricating oil also introduces uncertainties in the NG-based P M output. As shown in the next section, the shock tube is an ideal device to simulate diesel-engine combustion in terms of compression-ignition, direct injection, and consistently achieving relevant temperatures and pressures. The elimination of the diesel pilot and lubrication oil allows the direct observation of natural gas-based P M variations. In addition, the shock tube's quiescent high-temperature and high-pressure environment minimizes uncertain-ties in P M formation and oxidation due to fluctuating process conditions. However, the shock tube and diesel-engine dynamics also differ in several key respects. Shock tube ex-2.5. Particulate Matter Sampling from Shock Tubes 9 perimentat'ion involves a single injection and combustion event at a time, which does not resemble the continuous cycle of engine operation. The use of a diesel pilot or glow/spark to ignite the main natural gas fuel is also not applicable in the shock tube, where slightly higher than diesel-relevant temperatures are required to achieve auto-ignition of the test fuel. It is important, therefore, to accurately determine any NG-based P M correlations to aid further HPDI or other NG-fueled engine testing and research. 2.5 Particulate Matter Sampling from Shock Tubes Particulate matter sampling is performed routinely from engine emissionsKwhereby cylin-der dumping is immediately followed by sampling of the exhaust contents. The engine can be operated in a steady-state fashion, with the continously recorded emissions data representing an average of the numerous combustion events. However, due to the com-plicated variable interactions and flow dynamics within the engine, it is rarely possible to change exactly one operating parameter while keeping all the other conditions con-stant. Therefore it is very difficult to properly correlate the measured P M amounts with specific process variables. In the shock tube, however, the sampled P M quantities can be attributed directly to the input fuel. In this study, the possible relations between P M and methane-based combustion parameters will be investigated to support HPDI and other NG-based engine research and development. f 2.5.1 Shock Tube Dynamics Shock tube research in combustion applications such as ignition delay and pollutant formation has been generating increasing interest over the years. This is mainly due to its relative simplicity and versatility as a combustion device, as well as the ability to calculate various relevant combustion parameters with high accuracy. For example, measuring rapid temperature changes in internal combustion engines is a very difficult task, whereas temperatures in the shock, tube can be calculated from shock velocities that can be measured with high accuracy. However, it is important to understand the principle and theory of the shock tube apparatus, in order to understand the associated experimental procedure, results, and errors described in subsequent chapters, as well as to properly develop the experimental methodology needed for the P M sampling system. A shock tube is essentially a rigid cylinder (usually of constant cross-section) separated into two sections 2.5. Particulate Matter Sampling from Shock Tubes 10 by one or two diaphragms (acting as pressure barriers) mounted normal to the axis, as shown in Figure 2.4. It is usually also fitted with pressure and temperature measurement devices, as well as other relevant diagnostic accessories. A pressure difference is then applied across the high-pressure driver section and the low-pressure driven section. When the diaphragms eventually burst, a plane shock wave forms and travels into the driven section and a rarefaction fan goes into the driver section (Fig. 2.4b). The incident shock is a compression wave with locally supersonic speeds, and causes a pressure and temperature jump across its front [19]. A contact surface that separates the driver gas from the driven gas follows the incident shock wave and travels at a lower speed. At the same time, the initial rarefaction fan (series of expansion waves) reflects off the end wall and travels into the driven section. After the incident shock reflects off the end wall, it meets and stops the initial driven gas. The static high temperature and pressure reservoir generated behind the reflected shock wave is the main region of focus for all shock tube studies. For meaningful combustion studies, it is important for this experimental region to maintain a constant temperature and pressure environment, thus maximizing the experimental time period. This is achieved by obtaining a tailored interface condition, or where P 5 equals P6 in Fig. 2.4d. The tailored condition is achieved by using the appropriate species and amounts of gases in the driver and driven sections. The main principle involved is to attain a desired speed of sound (incident shock velocity) for the driver gas, in order for the shock wave to increase, the driven gas temperature and pressure to the target conditions [19]. This sound speed is calculated based on the gas constants and specific heat ratios of the driver gas constituents, and will be described in the experimental procedure. Since the oxidizer is air for the purposes of this study, helium (with relatively higher R and C p ) is used as the predominant part of the driver gas in order to obtain the required sound speed as calculated in Equation 2.1. a=y/iRT (2.1) When the pressure behind the reflected shock in the driven gas (P 5 ) is the same as that of the driver gas (Pe), no pressure fluctuations exist as the reflected shock travels across the contact surface. Therefore the experimental conditions will remain quiescent and undisturbed until the arrival of the reflected rarefaction fan significantly decays the pressure and temperature. However, if an inappropriate incident shock velocity is used, the reflected shock (after passing through the interface) will cause the driver gas pressure 2.5. Particulate Matter Sampling from Shock Tubes 11 (a) Driven gas Pi 3EZ Diaphragm Driver gas (b) Incident shock front Contact Rarefacrionfan. surface ^ — 1 ~ - * U 4 (c) 1* Reflected shock -flfront Reflected rarefaction fan P 5 " 5 P2=P 3 Figure 2.4: Shock Tube Principle [19] to be greater than the driven gas pressure, or vice versa. These result in undesirable phenomena called under-tailored and over-tailored interfaces, respectively. A n under-tailored interface causes the test pressure to rise steadily as the contact surface encroaches into the driven gas, while an over-tailored interface causes the experimental pressure to decrease due to the contact surface retreating from the experimental region. Therefore, the driver gas specific heat ratio must be carefully tuned to achieve the desired sound speed (thus negligible pressure difference) across the contact interface. By targeting the relevant temperatures and pressures, diesel engine-like combustion can be simulated in the shock tube and corresponding quantities of interest such as particulate matter emissions can be measured. 2.5. Particulate Matter Sampling from Shock Tubes 12 2.5.2 Particulate Matter Sampling Instrumentation Conventional engine emissions P M sampling usually involves direct mass measurements, such as the Tapered Element Oscillating Microbalance (TEOM) . If the P M mass is sufficient (tens of micrograms), fine collection filters are placed in the exhaust flow to trap all P M and subsequently weighed using sensitive scales. The P M concentration can be determined from the flow rate and collection time interval. For real-time measurements, the T E O M is a dedicated mass concentration instrument operating on Hooke's law. The mass of a Teflon-coated glass filter (fixed at the end of an oscillating tube) increases from the gradual P M deposition, which changes the natural frequency of the oscillation. Recent advancements in indirect mass measurement techniques have also been used in engine P M studies. For example, the Aethalometer uses an optical absorption method to determine the amount of black carbon (or soot) deposited on a quartz filter. Indirect mass measurements are advantageous for differentiating particular P M species based on certain chemical or physical properties. Due to the high amount of P M expected, diesel-engine exhaust is usually diluted before any of these measurements take place. Methane, however, is known to form little soot because of its unlikely molecular struc-tural transition to any soot precursor species. The expected P M mass from methane (especially non-premixed combustion) is extremely low. Therefore, particle sampling instrumentation in low concentration environments must be considered. For example, particles in the ambient atmosphere can be gathered gravitationally on a surface by sedimentation [3] [51], and the mass can be subsequently determined by gravimetric analysis. Black carbon amounts in urban areas have also been indirectly measured by links to other common combustion products (e.g. carbon monoxide) [4]. However, if continuous real-time black carbon measurements are required in relatively low levels (e.g. ambient atmosphere), the Aethalometer is often used. For example, diurnal patterns in black carbon concentrations are observed beside highways where truck (i.e. diesel-engine) exhaust peaks during daily traffic peaks [3]. Aethalometers have also been used to investigate indoor and outdoor B C sources for an occupied house [34], where contributions such as rush hour traffic, industrial plant emissions, cooking, and candle burning were accounted for. Additional studies on the B C level monitoring in workplace and vehicular settings have shown significant promise [10]. 2.5. Particulate Matter Sampling from Shock Tubes 13 Unfortunately, overall P M mass-related measurements are usually inadequate for determining the relative amounts of specific particle species. Unique properties (e.g. size) for the P M species of interest are typically used to measure their individual quantities [16]. For example, tuning collection filter porosities and materials for specific particle species can be used as a crude sizing technique. Inertial impactors are more commonly used directly in the particle flow stream to filter out the larger unwanted particles. Using the Stokes Number as the principle parameter, particles larger than a critical size will impact on the trap surface, while the rest will remain airborne and continue to the sampling instrument [36]. Hypersonic jets within the impactor are needed for nanoparticles. Cascade (or multi-stage) impactors are sometimes employed to achieve various particle size classifications by removing progressively smaller particles at each impaction stage. The Differential Mobil ity Particle Sizer (DMPS) is another popular particle sizing instrument in aerosol research. D M P S uses the principle of separating particle sizes by their electrical mobility. By applying specific electric field strengths in a Differential Mobil ity Analyzer (DMA) , the target-sized particles will be forced into a secondary flow stream [31]. A Condensation Nucleus Counter (CNC) is subsequently used to find the number concentration for each size range. Particles are artificially enlarged by alcohol condensation to enable optical detection. In addition to size, other signatures of the desired P M species can be utilized. For soot, its optical properties are used in an indirect technique in the Aethalometer. Due to the unique microstructure of aerosol black carbon atoms (as compared to other carbon forms), the electrons are mobile enough to absorb optical photons [14]. It has one of the largest broad-spectrum absorption cross-sections known and is strongly optically absorb-ing in the visible spectrum wavelengths. Therefore, the optical absorption measurement of 'blackness' can provide the basis of determining the amount of soot in the P M sam-ple. The blackness measurement was also found to be sensitive only to the amount of black carbon, and is insensitive to any extractable organic carbon or other optically non-absorbing aerosol species (except mineral dust). Therefore its visible light absorption may be interpreted directly in terms of B C mass [14]. Since there are no known significant biological, geological, or meteorological sources of aerosol black carbon, the measured B C quantities can be correctly attributed to combustion. 2.5. Particulate Matter Sampling from Shock Tubes 14 2.5.3 In-Situ PM Experiments in the Shock Tube Shock tubes experiments on particulate matter have been traditionally limited to in-situ studies of chemical reaction properties and species formation rates. Although there are relatively few studies on methane or gaseous fuels, the existing literature on in-flame particulate matter measurements are still useful to understand (see Table A . l ) . Alexiou and Will iams [1] [2] have studied soot induction times and formation rates using laser beam attenuation at various wavelengths, citing an Arrhenius dependence on shock temperature. Muller and Wit t ig [37] used an optical dispersion quotient method (based on the extinction of two laser beams) to observe the strong temperature dependence of soot induction times and soot volume fractions. Park and Appleton [39] found that soot oxidation rates, using laser light transmission, increases with increasing temperatures. Cadman and Denning [7] [8] also studied soot oxidation rates using laser beam attenuation, with no direct dependence on temperature or pressure cited. Kunz and Cadman [6] [33] also investigated in-situ soot particle emissions (using laser light absorption) from alcohol fuels, and found very small soot yields and possible correlations to the integrated light intensity. The above studies involve a wide range of shock temperatures, pressures, and non-methane fuels. The shock tubes are also equipped with optical- access to enable these optical-based measurements. In addition to laser light attenuation, chemical reactions and soot formation/emissions in the shock tube are also studied by more complex techniques such as time-of-flight mass spectrometry, laser schlieren densitometry, and laser induced incandescence [30]. As the more pertinent fuel to the current investigation, methane-based soot formation and particle studies have been conducted by Kellerer and Muller [27] [28]. Soot formation under fuel-rich premixed methane combustion were studied using Argon laser light ex-tinction/scattering at 488 nm, under pressures from 15 to 100 bar and temperatures from 1600 to.2100 K [28]. Particle diameters below 30 nm were observed at high pressures, as well as increased soot volume fractions with higher carbon concentrations. Soot yield (% of total initial carbon converted to soot) was found to be increasing with temperatures up to 1780 K and increasing with a smaller dependence on pressures up to 30 bar. In addition, growth and coagulation of soot particles under fuel-rich conditions with laser light extinction (Helium-Neon 633 nm) were also studied [27],under pressures from 10 to 60 bar and temperatures from 1500 to 2300 K. Particle diameters between 15 and 40 2.5. Particulate Matter Sampling from Shock Tubes 15 nm were measured, with the number density increasing strongly with volume fraction. Soot volume fraction was found to increase with carbon concentration, increase with temperatures up to 1700 K, but without any pressure dependence. However, the total emitted mass using soot yield and volume fraction results are difficult to calculate due to uncertainties associated with optical methods and oxidation tendencies. 2.5.4 Shock Tube Experiments of PM Emissions Although in-flame soot particle studies are non-intrusive, their local measurements cannot necessarily be directly translated to global amounts. Even with large amounts of reaction progress data, predictions of actual soot emissions from the shock tube are difficult due to the unpredictable nature of competing formation and oxidation reactions throughout the experimental region. As a result of the low volume fractions frequently observed in in-situ studies (~ 10~ 1 0), global particulate matter sampling from methane flames in a shock tube has not been previously attempted. However, literature on particle sampling methods from shock tube emissions is cited (Table A.l.) to aid the methodology development of the current methane flame investigation. Sidhu et al [44] studied particles extracted from premixed compressed natural gas (CNG) combustion at pressures from 20 to 27 bar and temperatures from 1000 to 1500 K. A nominal set of experimental conditions (mean temperature of 1150 K at 24 bar with an E Q R of 3), similar to those experienced by fuel during a diesel engine cycle, was selected. These conditions resulted in good ignition of the test sample with an ignition delay of 510 /xs. Scanning electron microscope (SEM) analysis showed dark gray and black aggregates made up of components less than 100 nm in diameter. The average P M mass collected from three tests was 390 fig with a particulate yield of 0.30% and a SOF of 38%. Although these findings were consistent with previous diesel and propanol combustion emissions under similar conditions, 100% P M collection efficiency was assumed without any supporting evidence. To evacuate the exhaust gases after each experiment, the exhaust valve to a sampling subsystem (attached to the driven section) is immediately opened. The particulate fraction of exhaust gases are trapped on high volume glass fiber filters while the volatile gases are captured in Tedlar sample bags (downstream of the filters). The filters are weighed before and after the test gravimetrically to determine the total P M mass. Even though the P M amounts enabled direct mass measurements, their extremely small percentages of the filter mass (~ 0.28%) resulted in a relatively 2.5. Particulate Matter Sampling from Shock Tubes 16 high variance of 24%; Wang and Cadman [49] [50] also performed shock tube extraction of particulate samples. Non-premixed combustion of liquid benzene sprays was used to detect the presence of Ceo in soot particles under a pressure and temperature of 2 bar and 2400 . K, respectively [49]. After each experiment, the shock tube contents were vented and pumped out. Particulates were trapped on high-pressure Whatman (Type 1) filters, and subsequent gravimetric analysis was used to determine the soot yields. Soot and P A H particles were again studied in the combustion of various other liquid fuel sprays, under pressures of 2-25 bar and temperatures of 1000-3000 K [50]. The shock tube contents were once again vented and pumped out immediately after each test, with the particulates being trapped on Whatman (Type 3) filters. The soot yield was also determined by weighing the filters before and after particle collection. Strong temperature and weak pressure dependences (within engine-relevant ranges) on soot emissions were reported in both cases. For example, soot yields for n-heptane ranges from 2 to 7 mg/g between 1000 and 1700 K. Although a particle collection efficiency of 90% was claimed in [50], no supporting evidence was given. The shock tube was simply cleaned by a cloth (before each test) and by high temperature, high oxygen blank tests (on a regular basis) to burn out remaining particle residues. Other existing shock tube particle extraction and sampling research includeBrezinsky of the University of Illinois at Chicago [5], where various sized particles of interest are collected. For larger sizes in the micron range, high-pressure Millipore filters were placed between the driven section and the sample vessel during the gas venting process, as shown in Appendix A. However, since soot particles in the nm range require high-resistance membrane filters, they are allowed to first settle on the shock tube walls. A n o-ring is placed around the circumference of a cylindrical rod so that the o-ring contacts the entire inner wall surface, as the rod travels from the diaphragm section to the end of the driven section. A Ziploc bag is used to collect the soot particles accumulated at the end of this cleaning process. Gravimetric analysis is used to determine the total P M mass for all sizes. It can be seen that this soot collection procedure makes it very difficult to esti-mate the collection efficiency, and will only be viable for high soot producing experiments. The current methane study differs from the available literature in several respects. Studies of gaseous species formation kinetics and emissions were not concerned with diaphragm 2.6. Conclusions 17 fragment interference and particle losses, with even less concerns of contamination con-trol. The relatively large quantities of P M generated (using liquid fuel sprays and pre-mixed gaseous fuels) also reduced the level of scrutiny required for particle losses and instrument detection issues. Although contamination control and particle losses were mentioned, it is not known what fractions of losses can be attributed to specific mech-anisms. The particle sampling paths and processes were also briefly mentioned without any elaboration of apparatus and/or procedural developments. In addition, most au-thors indicate clean and reproducible methods were used to initiate the shock wave, without any specific comments on diaphragm fragments and possible disturbances to experimental conditions. Instrument detection issues were explored even less, as direct mass measurements were used in the most relevant cases. Last but not least, effects of ignition variabilities in terms of time delay [21] and kernel location [47] were not taken into account, even though they clearly exist. 2.6 Conclusions The soot formation phenomenon in hydrocarbon combustion can be studied in its se-quence of stages: precursor formation, particle inception, particle growth, and oxidation. The fundamental chemical and physical processes are both kinetically and thermody-namically controlled, as well as being highly dependent on initial conditions. Numerical predictive modeling of soot formation is usually inadequate and thus experimental stud-ies provide the bulk of the reliable data. To realize the P M emissions reduction potential of NG-fueled compression-ignition engines, Westport's HPDI uses a diesel pilot to ignite the main N G charge. The P M contributions from multiple sources are difficult to sep-arate and therefore a shock tube is used to investigate the natural gas-based P M quantity. Although the shock tube is a common combustion device for in-situ P M formation studies, quantitative global P M emissions measurements still pose a significant chal-lenge, as evidenced by the lack of previous work in this area. Conventional engine P M emissions and atmospheric aerosol measurements are used to understand the particle differentiation methods based on size and light absorption properties, with the latter being most relevant for low volume fraction soot studies. Most of the existing shock tube research use premixed conditions and liquid fuels, as well as being focussed on chemical 2.6. Conclusions 18 kinetics, ignition characteristics, and gaseous species measurements. Existing in-flame soot measurements mostly involved laser-light attenuation techniques and reaction progress studies. In order to 'translate the local results to global quantities, relevant shock tube gas venting and sampling studies are used to understand the pertinent existing sampling methodology and its associated technical challenges. The reported soot yields from injection experiments are very small, and cannot be directly compared to the premixed results. In the absence of directly applicable real-time soot measurements in low concentration environments, various elements of the above research areas must be utilized and combined. However, the existing literature helps to identify the critical sampling system issues to overcome. The P M sampling system and methodology for the current investigation can thus be developed by systematically identifying and addressing each key technical chal-lenge, while managing the overall system objectives. Existing knowledge from engine P M sampling, shock tube soot and P M studies, as well as general soot properties should form a strong basis for this novel measurement system. After validating the sampling methodology, it is hoped that global soot emissions measurements from methane flames can be accurately obtained under diesel engine-relevant conditions (P ~ 30-40 bar, T ~ 1000-1200 K ) . Chapter 3 Development of Experimental Methodology 3.1 Introduction Particulate matter sampling from shock tube combustion is a relatively novel concept, and requires considerable theoretical and practical considerations from areas such as combustion, pollutant formation, fluid mechanics, heat transfer, as well as general aerosol and particle dynamics. These considerations are especially important for the types and amounts of fuel of interest in the current study, as the main challenge is to accurately measure the very small levels of P M expected (e.g. soot volume fractions of T O - 1 0 ) . The U B C shock tube facility has previously been used successfully in ignition delay studies and optical diagnostics, and has the potential to extend its capabilities to measure various pollutant emissions. This improvement will greatly enhance the shock tube's versatility as a tool in fundamental combustion research. However, the principle and operation of the shock tube presents several technical challenges in achieving the desired sampling system objectives outlined above, and these need to be addressed meticulously from theoretical-and practical perspectives. The critical experimental challenges are grouped into four main categories: obtaining a controlled experimental environment, controlling possible contamination sources, consid-eration of particle loss mechanisms, and careful interpretation of instrument detection issues. After an overview of the existing shock tube facility, each of these critical issues will be described in detail along with their corresponding solutions. As some of these 19 3.2. Description of Facility 20 technical challenges (as well as their solutions) are closely related and even overlap, they will be approached in parallel wherever possible. However, it is also important to make steady progress in each individuaHssue as to attribute resultant improvements to partic-ular modifications. As expected, implementing some of the solutions to these issues brings forth new problems, and although efforts are initially aimed at elimination of every prob-lem, minimization is usually the only practically achievable result. Finally, the overall P M sampling system as an end product will be summarized, along with the appropriate operational procedures developed. 3.2 Description of Facility The U B C shock tube is 7.37 m in length and 5.90 cm in inside diameter, separated into a 3.11 m driver section and a 4.26 m driven section by a removable double diaphragm section (see Figure 3.1). The tube is stainless steel (316 SS), with all fittings and connections designed for the purposes of shock tube studies. Experimental durations for this setup are in the range of 5-6 ms, with a wide range of experimental temperatures and pressures achievable behind the reflected shock. Typically, experiments under engine-relevant pressures are performed, with temperatures above the fuel-specific ignition limits. The double diaphragm setup (Figure 3.2) allows the pressure difference across each diaphragm to be maintained safely below its burst pressure, thus tolerating minor variabilities in material property and manufacturing processes. A vacuum pump is used to evacuate the shock tube prior to each experiment. The driven section is filled with the oxidizer (usually air), while the driver gas is composed of helium and air. The driven section is also fitted with dynamic pressure transducers to measure the incident shock passage. The shock tube accommodates both premixed fuel-air mixtures and non-premixed direct injection of gaseous fuels after shock reflection. In premixed experiments, the desired stoichiometric amount of fuel is added to the air in the driven section without attaching the optical section. In the non-premixed case, a solenoid type gaseous fuel injector is mounted to the end flange of the driven section (Figure A.4), and attached to the high-pressure gaseous fuel cylinder. The injector timing circuit is triggered by the passage of the incident shock, with the injection synchronized by the customized injector controller and driver. A detachable optical access section (35 cm in length) has been fabricated for the purposes of camera and laser diagnostics of 3.2. Description of Facility 21 the combustion zone, thus greatly increasing the amount of useful information from each experiment (Figure A.5). The optical access utilizes a separate section of the tube (carbon steel construction with the same inner diameter), with openings for insertion of quartz windows. The windows are properly sealed to withstand the temperature and pressure peaks during the combustion process. This setup allows light emissions to be transmitted, and the signal can be picked up by optical fibers placed on the window and amplified/filtered to recordable signals by a photomultiplier (Figure A.6). Signals from the pressure transducers, injector, and optical fibers are recorded on the data acquisition system and stored in a computer. A picture of the physical setup of the shock tube, including the double diaphragm system, injector connection, and the optical access section are shown in Figure 3.3. The detailed information on each apparatus and related experimental procedures will be described in the next chapter. This shock tube has been previously used primary to measure ignition delay of methane and associated additives, under premixed and injection environments. Schlieren imaging with a C C D camera has been performed through the optical access section, in order to capture snapshot images of the combustion process at different times after injection. More recently, a high-speed C M O S camera is added as a diagnostic tool to both increase the frame rate and allow a more detailed visualization of the entire combustion event. The primary goal of the particulate matter sampling system design is to create appro-priate add-on components to the existing shock tube facility, with minimal modifications to existing equipment. It should be compatible with present diagnostic tools, as well as foreseeable future gas sampling upgrades (e.g. N O x , gas chromatography analyzers) and particle analysis instruments (e.g. particle sizing, number concentrations). If possi-ble, the sampling system should also be versatile enough to allow investigations of other combustion parameter effects on shock tube P M emissions, such as fuel additives, equiva-lence ratio, and injection geometry. In order to accurately and correctly measure the low P M amounts expected from methane flames, detailed considerations of sampling system components relating to each critical issue identified above must be taken into account. 3.2. Description of Facility 22 Data Acquisition System Double Diaphragm • fi D Dynamic Pressure | -f V Sensor Vacuum Pump Fuel Driven Section n Static Pressure Sensor 6 1 Vacuum Sensor yy^ Helium Driver Section Figure 3.1: Shock Tube Apparatus Schematic [19] Figure 3.2: Double Diaphragm Setup [21] 3.3. Achieving Desired Experimental Conditions 23 Figure 3.3: Shock Tube Facility 3.3 Achieving Desired Experimental Conditions In order to correlate particulate matter emissions to various combustion parameters, the shock tube must be able to consistently achieve the desired experimental conditions (i.e. maintaining the target temperature and pressure throughout the experimental duration). Although theoretical adjustments can be made using compressible flow dynamics described in Chapter 2, the physical bursting process of diaphragms introduces certain practical issues that are not relevant in previous shock tube studies. In addition, it is important to avoid both direct and indirect interference of fragmented diaphragm materials with the P M measurement methods described later on. As a result, obtaining a controlled experimental environment involves the selection of suitable diaphragm materials and developing various associated fixes, with the evolution process becoming a substantial part of the overall experimental methodology. The UBC shock tube previously used stainless steel and aluminum sheet diaphragms with precisely machined grooves in the center to allow petals to open upon bursting. The remaining sheet thickness is representative of the burst pressure. This manufacturing method introduces various machining inconsistencies and defects, resulting in numerous unsuccessful experiments, in addition to the high cost of setting up a new batch. To drastically reduce the material cost, various shim stock materials and thicknesses have 3.3. Achieving Desired Experimental Conditions 24 been tested and found to be very consistent in their burst pressure and throttling loss coefficients (Appendix B). In particular, standard thicknesses of stainless and carbon steel shim can achieve a wide range of experimental pressures without machining. The stainless type was initially used because it is relatively stronger. It also bursts into small fragments, causing minor disturbances as it is blown downstream into the experimental region. However, in an effort to prevent P M instrument signal contamination from the diaphragm particles (see Contamination Control section), magnetic filtering was used to preferentially trap diaphragm materials while allowing soot particles to enter the mea-surement device. Although stressed stainless steel particles are slightly magnetic, carbon steel shim diaphragms are used to ensure maximum magnetic capture efficiency. The main drawback of this material is shown in Figure B.2, where the entire cross-sectional area breaks off at the burst point, causing a large projectile to travel rapidly downstream. In addition to scratching the delicate quartz windows on the test section, this large piece of diaphragm often results in a large and sudden pressure spike within the experimental region (Figure B.4), causing significant and unpredictable local pressure and tempera-ture fluctuations. Since soot formation is strongly dependent on local conditions, the stoichiometry at which these pressure spikes occur will drastically affect the subsequent P M formation and oxidation processes, resulting in unpredictable emissions results. ' 3.3.1 Elimination of Pressure Disturbances Since the main drawback of steel shim diaphragms is large local pressure disturbances in the experimental region during the experimental duration, it is attempted to eliminate this effect by trapping diaphragm fragments prior to their entry into the controlled experimental environment. The main approach used is the insertion of a cylindrical ring with an intertwined wire mesh, with several versions shown in Figure B.5. The ring is supported by a vertical steel bar through the shock tube, placed approximately 6 cm from the first diaphragm to allow for bulging. The trap is located as close to the diaphragm section as possible to minimize shock formation disturbances due to the slight decrease in cross-sectional flow area. It has. been found that although large pieces are routinely trapped, they are occasionally cut by the wire 'net' into smaller pieces and continue to fly into the test section and affect its conditions. The use of a finer mesh prevents this problem, but significantly decreases the flow area and lead to unpredictable test conditions. Furthermore, the trap introduces variabilities in diaphragm fragment 3.4. Contamination Control 25 breakage and alignment on the incident surface, thus causing varying degrees of throttling loss and increasing the uncertainties in experimental conditions. Therefore additional fragment trapping efforts to eliminate pressure disturbances were not very practical and thus were subsequently abandoned. Instead of minimizing the diaphragm fragment disturbances on experimental conditions, a new material is needed to achieve a better compromise between experimental consistency and P M measurement interference. To maintain the burst pressure consistency, another ready-made sheet material with var-ious standard thicknesses is attempted. Lexan (polycarbonate) plastic is chosen for its potential in minimizing optical P M measurement signal interference, since its fragmented particles will not absorb significant amounts of light. Lexan diaphragms also fragment (Figure B.3) into very small pieces (with slower velocities) that do not produce large pres-sure and temperature disturbances in the experimental region. It is also the diaphragm material of choice in other shock tube research facilities [11] [24]. Detailed burst pressure and throttling loss coefficient testing information for all diaphragm materials discussed above are summarized in Appendix B. 3.4 Contamination Control Due to the small amounts of particulate matter expected, contamination control is an extremely important and challenging system design consideration, and therefore must be. carefully developed independently as well as in conjunction with particle loss and instrument detection efforts. The investigation and remediation of possible contamina-tion sources and materials can be most efficiently approached in a systematic manner. As a result, the design challenge of contamination control is separated into four areas: contamination detection, pre-experiment control, post-experiment control, and particle analysis. In each area, the objective is to identify the sources and type of possible for-eign materials, and then finding appropriate solutions to mitigate the problem. It is also important to note that the procedures and methods outlined below were not necessarily performed in the same sequential order, since some forms of contamination can affect more than one area while other forms can be tackled concurrently. For example, particle analysis is an onging procedure with its results used in the development of various control areas. Finally, it is also important to address the relevance of the contamination to the 3.4. Contamination Control 26 desired quantity to be measured, and to develop prevention and minimization methods accordingly. 3.4.1 Contamination Detection Contamination from the shock tube is detected under three different conditions using various available diagnostic tools. It is important to be vigilant in analyzing the clues from these results, in order to find all possible contamination sources later on. First of all, contamination without any shock wave or fuel combustion is searched. Blank tests (with a shock wave but without any input fuel) that simulate actual test conditions are then performed to check the baseline of the existing facility, hence the degree of foreign black carbon (BC) contamination. Finally, sample experiments are run with actual fuel injection to search for further clues between light emission characteristics and existence of foreign substances. Although the black carbon results using the Aethalometer will be shown as part of the contamination evidence, the detailed data analysis procedure developed for shock tube sampling purposes will be provided in the upcoming Instrument Detection section. Details of sampling bags and the associated apparatus/procedure can be found in Chapter 4, and thus will not be elaborated here. Non-Shock Tests In the absence of a shock wave, possible particle contamination in the shock tube can be checked by observing differences in black carbon amounts using different filling pro-cedures. Helium is used in this test since it is the predominant component of the shock tube exhaust gas sample. For additional procedural consistency, the volume of gas used for these detection experiments are very similar (observed by the degree of bag bulging). When the gas is first filled into the shock tube (without diaphragms), and then imme-diately dumped into the sample bag via the exhaust port valve, the black carbon mass profile obtained is shown in Figure C . l (details of the applicable mass calculations are shown in Appendix, E). However, if the gas is filled from the bottle directly into the bag (with similar configurations and flow rates as the shock tube dumping method), the mass increments per timebase period are significantly lower (Figure C.2). Although the amount of discrepancy between these two results depends on factors such as the degree of tube cleaning, there is consistently more black carbon after the gas has passed through the shock tube. Therefore any contact with the inner shock tube surface seems to introduce 3.4. Contamination Control 27 additional particles into the gas stream. B l a n k Tests Blank tests are performed to utilize the available diagnostic tools to detect contami-nation, as actual experimental conditions and shock wave dynamics are achieved. In addition, blank tests are important in establishing appropriate baseline black carbon amounts, before results from combustion experiments can be properly interpreted. Table 3.1 shows the blank test results superimposed on Figures C . l and C.2. A l l three graphs are plotted from post-processed Aethalometer data (resulting in black carbon mass accumulation against the sampling time), the detailed of which will be described later on. This increase above both non-shock filling methods is representative of all blank test results, although the magnitude of the increase can vary depending on the particular cleaning procedure used. Since the blank tests are giving consistently more black carbon than merely filling the tube (with all other conditions kept constant), the momentum of the shock wave seems to be entraining possible particle deposits on the tube wall into the main gas flow, thus resulting in the instrument picking up additional black carbon signal. Tab le 3.1: Contamination Detection Detection Method BC Level (ng/min) Bypassing Tube 0.5 Filling Tube 2.5 Blank Test 8.5 Note: representative mass increment values shown Light emissions from blank tests provide more evidence of contamination. A n example of the broad band photomultiplier signal during a blank test is shown in Figure C.3. It can be seen that there is a distinct rise in the signal above the background sensitivity level, during the high temperature and pressure experimental duration (after shock reflection). This seems to indicate that there, is foreign material in the tube, when carried by the shock wave into the end of the driven section, ignites under the experimental conditions. Similar blank tests with only inert gases (nitrogen and helium in the driven and driver sections, respectively) also show light signals above the noise levels (Figure C.4, calibrated to the same sensitity as Figure C.3). Nitrogen is.used (as the driven gas) in the inert blank 3.4. Contamination Control 28 tests to maintain the tailored condition. The amount of light detected in the presence of only inert gases suggests possible leakage of air into the tube during test preparation, and/or insufficient evacuation by the vacuum pump. Since both types of blank tests produce non-negligible amounts of light, the contamination sources must be located and controlled. C o m b u s t i o n Tests To investigate possible contributions and interference of the detected foreign particle contamination to the actual combustion process, light emission analysis and high-speed C M O S camera visualization is performed through the optical section. From the pres-sure and light signals of a representative methane injection experiment (Figure 3.4), it can be seen that the particle ignition process occurs immediately after shock reflection and precedes the main fuel ignition, and has very distinct light emission characteristics. Any contamination particles present in the shock tube are likely to ignite during the favourable conditions established immediately after shock reflection. As shown in Figure 3.5, representative frames (pre-injection, fuel ignition, and flame propagation) from the camera video also verify the existence of foreign particles. Initial bright spots are clearly visible throughout the field of view in all cases, and is clearly distinguishable from the much more brighter and broader fuel burning. In addition to sampling interference caused by these particles, their ignition and combustion can also contribute enough quantities of black carbon to contaminate the desired measurement (i.e. those resulting from the fuel alone). More importantly, particle ignition processes can have complex interactions with fuel ignition and soot production mechanisms; the ramifications of which will be difficult to account for in the total black carbon mass results. 3.4.2 Pre-Experiment Control Once contamination has been detected, it is necessary to locate the source and try to minimize or eliminate it. The efforts spent on controlling each form of contamination de-pend on factors such as cost, technical feasibility, convenience of implementation, as well as compatibility with existing experimental procedures. Pre-experiment contamination control includes all control steps prior to the actual shock initiation. The two main areas of concern are preventing contamination sources from entering the tube and removing existing contamination on the tube walls. Since it is far more effective to eliminate con-3.4. Contamination Control 29 Time (ms) Figure 3.4: Light and Pressure Signals from Combustion Test (20 bar, 1200 K ) tamination by preventing its initial entry into the tube, significantly more efforts have been spent on pre-experiment contamination control compared to diaphragm contami-nation and post-experiment control methods. C lean ing M e t h o d s As a result of numerous previous premixed and injection experiments in the shock tube, soot and other particles have been collecting on the inner tube surface. Small particles such as soot can adhere tenaciously through a combination of physical attraction, chemical bonds, and mechanical stresses [45]. Table 3.1 shows that some particle deposits are able to be pulled loose by the shock wave and subsequently combust under the experimental conditions. Therefore attempts were first aimed at thoroughly cleaning the tube to a pristine environment, while continuously monitoring the blank test black carbon levels. Some of these were one-time procedures while others were routinely employed in the eventual P M sampling methodology (described later on). Table 3.2 3.4. Contamination Control Figure 3.5: Frames of Pre-injection, Fuel Ignition, and Flame Propagation gives a summary of blank test results after the various cleaning procedures described below, using total black carbon masses converted from similar sample gas volumes. As a series of blank tests under similar conditions do not always produce consistent results, only representative and approximate values are shown to track the general progress and effectiveness of the various contamination control steps undertaken. Figure C.5 shows an example of the need for proper tube cleaning, where there is a strik-ing difference between contaminated (end of driven section) and clean (driver section) surfaces. Scales of dark foreign material (possibly soot) build-up are clearly visible at the end of the driven section, where most of the combustion event takes place. To check for possible contributions of these materials to the blank test contamination, the end portion of the driven section was reversed so that it was not exposed to the experimental conditions. Since the subsequent blank tests still showed contamination, physical tube cleaning was inevitable and the least intrusive method was attempted first. High 3.4. Contamination Control 31 temperature blank tests (with 50% of the calculated driven gas filled using argon instead of air) were performed to burn out the dark material build-up in the combustion region. In addition, high temperature (with low pressure) burnout of the entire tube was also attempted with hydrogen as the driven gas, where the flame propagation was expected to reach various surface crevices. However, subsequent blank tests with the usual driven gas showed no obvious reductions in light emissions and black carbon levels. Chemical cleaning methods were attempted next, using a carbon-removing solvent (Crystal Simple Green). Acetone was briefly used initially but abandoned due to its highly flammable nature. Clean cotton towels with high surface strand density and low fiber content were soaked in the solvent and water solution, wrapped around the end of a telescoping rod, and pulled tightly through the entire tube with vigorous oscillating motion throughout each section (Figure C.6a). After rinsing the tube with the towel soaked in water, it was allowed to dry before blank tests were performed. To allow extra time to dissolve large carbon deposits, the shock tube was also filled completely with the solvent solution for three days, and then rinsed with towels. Since the towel surface was not noticeably dark after passing through the tube, mechanical cleaning methods with larger surface forces on the tube walls were attempted next. To provide thorough surface area coverage, circular brushes with high bristle densities were used, with diameters slightly larger than the shock tube in order to ensure bending and significant contact forces. Brushes were pulled through the tube with a rope (Figure C.6b) several times, where the scaled layers in the combustion region were clearly reduced and fine particles were visible on the bottom of the tube. The carbon steel brush (compared to nylon and stainless steel) seemed to produce the highest scraping force and pulled the most scaled material off the walls. Brush cleaning were also combined with subsequent solvent cleaning in order to dissolve the particles more effectively after they have been loosened from surface crevices. Table 3.2 shows that the above cleaning methods are effective to a certain extent, however the lower blank test levels cannot be consistently achieved as new particles were gradually deposited from subsequent experiments. To increase the duration and probability of bristle-particle contact, nylon brushes were attached to | " hollow stainless steel tubing, which was clamped to a high-speed (~ 500 rpm) \ " dril l, as shown in Figure C.7. To minimize eccentricity of the long shaft, the end was passed through steel diaphragms with center holes punched out and clamped to the end flange. The entire shock tube was cleaned in three separate sections (driven section consists of two 3.4. Contamination Control 32 component tubes), with an emphasis on.the more contaminated driven section. A steady flow (~ 2-3 L P M ) of filtered water was fed through the top of each slightly tilted tube section (to maintain an accumulation of ~ | of the tube diameter), in order to wash away any loosened particles as well as provide general lubrication. Approximately 10 minutes of total cleaning time was used for each section, starting at the middle of the section (to ensure water contact) and pulling the brush very slowly to the end, and then repeating the procedure after turning the tube around. Each section was then blown dry with the air-jet cleaning tool (described later on). Steel brushes were not used in the drill-cleaning procedure to avoid large friction forces and surface scratching. Table 3.2 seems to suggest that there were existing deposits not reachable by the above techniques, thus requiring more drastic measures to access these particles. By inspection, most of the tube surface was far from being smooth, with visible scratches and distinct peaks and valleys. Since previously deposited soot particles (10-1000 nm) can easily be trapped in the surface crevices (with depths of the original tube surface roughness plus any diaphragm fragment scratches), it was important to gauge the extent of the existing surface roughness for the feasibility of further surface treatments. Analysis of the tube surface revealed groove depths of.0.02-0.03" (0.5-0.75 mm), thus limiting the amount of particle access from the previous cleaning methods. Micro-honing was therefore performed on all tube sections to remove 0.015 inches of material in all directions, using a set of circular honing stones and slightly abrasive paste [12]. The resulting R M S surface finish was in the 100 nm range, significantly limiting the ability of the particles to hide in the ridges. Subsequent mechanical polishing/buffing of the surface was not performed since it uses an oily substance and does not remove any additional materials. Blank tests after cleaning out the powdered scrappings from the honing procedure, however, showed similar black carbon levels as before. To further ensure a particle-free inner tube surface immediately prior to each exper-iment, all residual particles remaining after cleaning need to be efficiently removed. This includes particles not accessible by the cleaning tool as well as particles introduced during the cleaning procedures (e.g. towel fibers). Since shock waves are known to be very effective in entraining particles from surface crevices (Table 3.1), gas jets exiting at supersonic speeds were impinged around the inner tube circumference to remove any attached particles. The threshold pressure ratio required increases with nozzle-surface 3.4. Contamination Control 33 spacing and horizontal translation speed [45]. To maintain a reasonable pressure ratio and translation speed (in terms of cleaning gas quantity required), the nozzle to surface gap (H) was minimized (H /D = 1.5). Each hole has a diameter (D) of 2 mm, based on dril l bit' size constraints. Pressure ratios and axial speeds of approximately 5 and 15 cm/s respectively were used, resulting in air consumptions at 150-200 psi. To increase the cleaning effectiveness, a total of 10 evenly spaced holes were drilled into the chamfered circumferential edge of an aluminum cylinder, each directed at an angle of. 60 degrees to the surface to push the entrained particles in the axial direction and eventually out of the tube (Figure C.8). A positive pressure environment also needs to maintained upstream of the gas jets in the tube (after clamping the diaphragm section) to prevent lab air entry. Improvements due to the gas jet cleaning can be seen by inspection of the tube surface, where the amount of towel fibers and diaphragm fragments were clearly reduced. This supersonic air-jet cleaning was also effective in drying the tube surface after solvent cleaning. i In addition to tube cleaning, all tubing, fittings, valves, and associated plumbing con-nected to the shock tube (as shown in Figure 3.1) need to be initially cleaned to remove previous deposits, in order to prevent them from being entrained by inlet gases. This is especially important at the various valve and tubing exit locations where thermophoretic losses (as the pressurized exhaust gases are rapidly vented to atmospheric conditions) re-sult in visibly dark particle deposits. Most of these components were cleaned and rinsed (using solvent and water, respectively) with soaked paper towels wrapped around thin wires. Small circular nylon brushes were used to loosen particles (prior to cleaning) on the more accessible contaminated surfaces. In addition, various sections of old copper tubing on the main venting line (with very black interior surfaces) were replaced by new stainless steel tubing. Finally, the end plate on the driven section, with dark patches from prolonged exposure to the combustion region, was scrubbed with small stainless steel brushes in the presence of the solvent. Pa r t i c l e E n t r y P r e v e n t i o n To complement the various cleaning methods developed, efforts to prevent particle entry into need to be taken to achieve thorough pre-experiment contamination con-trol. The importance of ensuring that the post-cleaning environment is uncontaminated 3.4. Contamination Control 34 Tab le 3.2: Summary of Pre-Experiment Control Procedures Pre-Experiment Control Procedure Subsequent Blank Test Summary BC Mass (ng) Temperature (K) Pressure (bar) Before Cleaning 580 -620 1200 20 Reversed Driven Section 1730 -1770 1600 10 High Temperature Burnout (50% Ar in driven) 530 -570 1200 20 High Temperature Burnout (H2 in driven) 395 •435 1600 10 Water (driven section) 480 -520 1450 24 Acetone, Simple Green (driven section) 465 -505 1200 20 Simple Green, Brushing (driven section) 250 -290 1200 22 Soak Tube with Simple Green Solution 235- 275 1200 22 Drill Cleaning with Rotating Brush 1610- 1650 1200 30 Tube Honing 940 -980 1200 30 Fittings, Valves, Tubing Cleaning 1660 • 1700 1200 30 Note: masses are calculated based on normalized volume of 350 L is clear from Table 3.2, where post-cleaning blank test results were quite random and did not show a consistent downward trend. This seems to suggest that a pristine post-cleaning environment was difficult to achieve and maintain, thus particle entry prevention methods during and after cleaning need to be carefully considered. The two main possible contamination sources are room air and pressurized gas bottles. Preliminary tests using a C P C (without any size distribution by a D M A ) are shown in Table C . l , and confirmed previous evidence of particle deposits on the tube walls. It also shows the need to prevent ambient lab air from entering the tube, where dust and other atmospheric particles can be sources of ignition in Figure 3.5. A positive pressure environment is therefore impor-tant in all cleaning procedures, as well as sealing all cleaned sections promptly. Table C . l also indicates the need to control particle entry from inlet gases (e.g. air, helium, fuel), since pressurized gas cylinders can contain some particles. As a result, all gas lines entering the shock tube were fitted with 15 micron sintered filters (to match those on the fuel inlet line) to trap particles from these sources. 3.4.3 P o s t - E x p e r i m e n t C o n t r o l Post-experiment control involves black carbon signal contamination from the shock tube exit to the Aethalometer. This consists mainly of non-BC particles produced from the 3.4. Contamination Control 35 experiment (e.g. diaphragm particles during bursting) and particles remaining after pre-experiment control steps. Although the detection of this contamination will be explained in a later section, its presence will be assumed for the current discussion of elimination techniques. Three main post-experiment control strategies are developed, depending on the diaphragm material and particle size. i Magnetic Fil ter ing The bursting of stainless and carbon steel shim diaphragms can produce sufficiently small fragments that can stay airborne and subsequently be carried by the rapid venting process into the sample bag. Some of these particles are aerodynamically shaped.(e.g. delta wings) to easily become suspended in a fast moving flow. To avoid black carbon signal contamination in the Aethalometer (as shown in Table 3.1), these fragments must be efficiently removed from the gas sample flow without disrupting the path of the soot particles. Since stressed stainless steel is slightly magnetic, the bursting process induces an ideal property with which to capture shim diaphragm fragments immediately before the Aethalometer inlet. To increase capture efficiency [54], strands of permanent magnets were separated at small distances to create high magnetic flux fields. These disk magnets were strung on a thin wire and aligned with the same polarity sides facing, in order to force the small separation distances. The magnet diameters were chosen to be slightly smaller than the stainless steel tubing in which they are placed, thus forcing the particle-containing gas flow to make several passes through the relatively large surface areas (see Figure C.9). The tubing containing the magnetic filter was also larger than the rest of the sample line to enhance capture by decreasing flow velocities. Results of applying this control technique to experiments using steel shim diaphragms can be seen in Table 3.3, where the exhaust gas from the same experiment initially bypasses and then flows through the magnetic filter, before entering the Aethalometer. Particle Impactor When Lexan was used as the diaphragm material, a new filtering method was necessary to remove its fragments. Due to the difference in bursting characteristics (Figure B.3), Lexan fragments into finer and lighter pieces, resulting in larger numbers of particles escaping during the tube venting process. T E M analysis (discussed later) also showed that some of these particles can approach the size range of soot agglomerates 3.4. Contamination Control 36 and introduce complications in its control. A new separation method was needed to trap this contamination while allowing soot to pass through freely. As discussed in Chapter 2, particle impactors are widely used in aerosol research as size separation instruments, by utilizing various particle and host gas stream properties. For small-sized particles, impactors based on the inertial separation phenomenon are commonly used (see Appendix C for the detailed impactor theory). The particle size cutoff attainable by inertial impactors is related to the gas jet velocity, - where subsonic jets typically filter down to submicron sized particles while hypersonic jets (accelerated using nozzle configurations to compressible regimes) can reach the nanometer range [36], However, designing for submicron (especially nanoparticles) cutpoints is technically difficult, since small orifices are required in combination with the availability of high pressure drops of the host fluid flow [18]. Slight random variations in the particle velocities (by following different streamlines) can also lead to significant deviations in the ideal efficiency curve characteristics. Preliminary attempts using a conventional low-pressure particle impactor (2.5 tim cutoff, 3-4 L P M flow rate) did not result in noticeable reductions of black carbon quantities (see Table 3.3), while diaphragm particles are clearly found in the sample gas flow (see S E M section below). Since some of the Lexan particles can be as small as hundreds of nanometers in size, hypersonic impactors were likely required for interception. Therefore, a customized hypersonic inertial impactor (with small nozzle orifices) was developed to utilize the high pressure drop at the shock tube exit valve to achieve high flow velocities. Initial results from a single stage, single nozzle | " hypersonic impactor attached directly downstream of the valve (Figure C . l l ) during the venting process showed significant promise (Table 3.3), although its effectiveness was reduced in subsequent blank experiments. This could be due to the progressively' smoother surface of the collector plate, as well as variations in streamline paths [36]. In all experiments, Lexan-like (white) particles were clearly visible on the collector plate. In some instances brown (possibly burnt Lexan) and black (possibly soot) particles were also present in various quantities. The sizes of the particle dots (directly downstream of the nozzles) on the plate also seemed to be quite variable. To improve the impaction efficiency at the design cutoff size (initially set at 300 nm), a three-stage impactor was constructed where each stage consists of a large nozzle and removes a portion of the unwanted particle size ranges. As shown in Figure C.12, this impactor also attaches to the shock 3.4. Contamination Control 37 tube exit valve and each stage consists of a brass disk inside a tee-fitting, with a small amount (< 10% of total volume) of bleed flow past the disks. Each collector stage has dn of jj" and L of half the tube diameter. The flow area gradually increases to compensate for the decreasing pressure drop across each subsequent disk, in order to achieve consistent high velocities. Table 3.3 gives a summary of the results using this three-stage hypersonic impactor, which did not seem to improve upon the previous single-stage version. Inspection of the impactor plate surfaces showed similar variability in the color and size of the particle dots. Furthermore, the amount of particles gradually decreased from the first plate surface to the last plate surface, and was consistent in all the trial experiments. This seemed to indicate that the pressure drops in the second and third stages were insufficient to provide the flow-velocities needed for their respective cutoff sizes. To further improve the impaction design and efficiency, a new single stage impactor with multiple nozzles (in concentric circles) was designed without a bleed flow. This alleviates the pressure drop deficiency by maximizing flow velocites through each nozzle. The resulting improvements in the number of impacted particles faciliated larger (1" diameter) nozzle and collector plate areas, thus allowing even more nozzles to operate at hypersonic conditions. The increase from the | " shock tube vent tubing to an intermediate i " tubing and then to the 1" irnpactor.nozzle plate allows a gradual flow velocity reduction in order to distribute the flow rates over each nozzle evenly (see Figure C . l l ) . The brass collector plate and nozzle plate was separated with a ring of thin wire (L — 0.027"). As shown in Figure C.13, the nozzle holes were aligned in different concentric circles than the collector holes, thus allowing the smaller particles to make the turn and continue downstream into the bag to be sampled. Due to the close vicinity of the two plates, hard solid particles below the cutoff size can sometimes bounce off the collector plate surface, thus decreasing the collection efficiency [18]. A thin layer of surface coating fluid (motor grease) was therefore applied evenly on the collector plate before each experiment to minimize particle bounce. Results from this multiple nozzle impactor are also shown in Table 3.3, where the reduction in the blank test levels can be consistently achieved by using the particle settling technique (described below) in conjunction. The size and darkness of the particle dots on the collector plate were still variable from run to run and did not seem to correlate with Aethalometer- results. This indicates that some soot particles could conceivably be impacting the plate, while some fine Lexan particles were 3.4. Contamination Control 38 still able to make the turn and continue downstream into the sample bag. Linear white flow patterns on the collector plate (from each nozzle plate hole to the closest, collector plate hole) give further evidence that significant portions of the diaphragm fragments were not trapped by the plate. Therefore, the impactor plate deposits can only be used as a qualitative correction technique with respect to the amount of combustion P M in the gas sample. Par t i c le Se t t l i ng Due to the inconsistent results from the various impactor designs, a more reliable di-aphragm particle filtering method is needed. The difference in Lexan and soot particle sizes and masses results in different settling velocities in a quiescent fluid, and thus can be utilized by venting the exhaust gas with a time delay after the experiment. The amount of settling time is calculated using the diaphragm and soot particle sizes from the T E M analysis. Since the inner diameter of the tube is relatively small, the shock tube contents should be settled long enough for the Lexan particles to contact the wall, but before significant settling and diffusional losses of soot particles occur. Preliminary blank exper-iment trials with various settling times (in conjunction with the impactor) are shown in Table 3.3, where the sensed post-settling black carbon levels are consistently lower. It is found that 60 minutes is sufficient to settle out a significant portion of Lexan diaphragm particles. Settling results are compared by withdrawing gas samples from the same ex-periment at two different times, due to natural run to run P M variations. Since any unnecessary additional settling time will likely result in undesirable soot particle losses by various mechanisms discussed later, further settling studies on soot particle losses will be needed to find the optimum tradeoff time. B a g C o n t a m i n a t i o n In addition to controlling the release of diaphragm particles from the shock tube, it is also important to prevent particle accumulation in the sample bag from entering the Aethalometer. Since the sample gas residence time and the difference in inlet and exit flow velocities are relatively large compared to the associated tubing, the bag can easily act as a reservoir for contamination. Therefore a procedure is developed to ensure an acceptable background black carbon level exists before shock tube gas samples are dumped into the bag. This post-experiment control method involves filling the bag directly with clean 3.4. Contamination Control 39 Table 3.3: Summary of Post-Experiment Control Procedures Post-Experiment Control Procedure BC Mass (ng) Blank Test Summary Temperature (K) Pressure (bar) Without Magnetic Filtering 795 - 835 1020 40 With Magnetic Filtering 420 - 460 1020 40 . Without Impactor 940 - 980 1350 20 Conventional Low-Pressure Impactor 935 - 975 1350 18 1/2" Single-Stage Impactor 345 - 385 1300 30 Multi-Stage Impactor 670-710 1300 30 1" Multi-Nozzle Impactor 330 - 370 1300 30 Settling Time: 0 min. vs 60 min. 575 vs 95 1300 30 Settling Time: 10 min. vs 70 min. 315 vs 135 1300 30 Settling Time: 0 min. vs 130 min. 715 vs 90 1300 30 Note: masses are calculated based on normalized volume of 350 L bottled gas, under similar flow rates to simulate the shock tube venting process. The filled bag contents are then immediately pumped out to ensure any airborne particles inside are removed. The bag is then filled with bottled gas again in the same manner, and sampled by the Aethalemeter to check for any significant contamination above the 'clean' background as shown in Figure C.2. 3.4.4 P a r t i c l e Ident i f ica t ion Particle identification involves microscopy and elemental analysis of particles to trace possible contamination sources. Although the results are grouped in this section, it is an ongoing procedure of contamination control and is complementary to the discussion in previous subsections. Surface wipe tests using fine silica filters and adhesive stickers on the experimental region are initially performed to locate possible tube contamination sources. Aethalometer quartz tape samples are then analyzed to identify any measurement signal contamination. Finally, lacy carbon grids are used to analyze diaphragm and soot particle sizes in order to develop the impaction and settling methods described above. The grids will also be analyzed to confirm contamination sources in the gas sample immediately upstream of the Aethalometer, as well as to compare particles found from the previous 3.4. Contamination Control 40 microscopy analysis. In each case, the samples are analyzed with the scanning electron microscope (SEM) or the transmission electron microscope (TEM) to pick out individual particles. Detailed elemental analysis using energy dispersive x-ray (EDX) is subsequently performed on particles of interest, in an effort to determine their composition and origin. Particle sizes and shapes are noted to aid in the design of appropriate filtering methods, as sometimes it is unrealistic to completely avoid certain forms of contamination. Although representative pictures and elemental.charts will be shown here to facilitate the current discussion, the detailed microscopy results are included in Appendix C, along with a summary of various materials' elemental makeup in Table C.2 for identification purposes. S E M A n a l y s i s To locate pre-experiment contamination sources, small circular silica filters are used to swab the experimental tube wall region after blank experiments. Slightly adhesive stickers are also used to capture visible foreign particles in the vicinity. S E M / E D X analysis clearly shows several distinct particle types. Figures 3.6 and 3.7 show relatively large (~ 200 jum) oval-shaped rubber gasket (Figure 3.2) fragments, likely cut off by the bursting action of diaphragms. Silicon and oxygen contained in the elemental breakdown are from the silica filter background. Since these carbon particles can ignite and affect the P M results, all rubber gaskets are replaced with Teflon to eliminate direct carbon contamination. Figures C.14 and C.15 show strands of dust particles (~ 200 /j,m) which likely entered from lab air prior to the experiment. Strands of fiber (possibly from cleaning towels) consisting mostly of carbon are also present in some samples. The discovery of dust and towel particles led to the aforementioned development of positive-pressure and post-towel cleaning techniques, respectively. To search for direct contamination to the black carbon measurement signals, sections of the Aethalometer quartz tape (after sampling from a blank run) are cut off and analyzed. Figures 3.8 and 3.9 show some small (~ 2 tim) stainless steel particles with a variety of shapes scattered on the relatively dense quartz fiber matrix. These particles are brighter than the gasket particles due to their composition of heavier elements. They are most probably from the bursting of the initial stainless steel diaphragms, and subsequently remained suspended in the gas flow. Since resultant artifacts in the total black carbon amounts (see Instrument Detection section) will be introduced, carbon 3.4. Contamination Control 41 177.9|im WD14 6mm 2 0 . 0 k V x l h O F igu re 3.6: Rubber Gasket Particle steel diaphragms are initially used as a substitution (in conjunction with magnetic filtering) before switching to Lexan. To detect potential Lexan diaphragm particle contamination in the black carbon mass from a blank experiment (from Table 3 .3 ) , S E M analysis was performed on the lacy car-bon T E M grids (described below). This procedure also checks the effectiveness of the first hypersonic impactor design (to aid in the development of subsequent versions), as well as to confirm the types of contamination particles previously found. A summary of representative particles found on a blank experiment grid (after passing through an impactor) and a sooty experiment grid are shown in Tables C . 3 and C . 4 . The background composition of the grids is first noted by pointing the S E M beam at the laces (without attached particles) between the main grid corners. Due to the carbon stubs used to hold the grids, the E D X analysis contains predominantly carbon, and the particle composi-tions are obtained by subsequently ignoring the carbon component. The main copper grids also interference in the elemental analysis, including those of the background laces (as the high-energy beam detects the nearby grid boundaries). In addition, hydrogen is usually too light for E D X detection. Therefore it is extremely difficult to obtain accurate element percentages without eliminating interferences and applying suitable beam cor-rection factors, especially for the lighter elements (i.e. differentiating soot and Lexan). Table C . 3 shows possible dust and Lexan particles are captured (see procedure in the 3.4. Contamination Control 42 kCts 3<H 20 H 10-E lem ent Concentrat ion Carbon 24.86 wt% Oxygen 53.50 wt% Sodium 0.73 wt% Magnesium 0.35 wt% Alum inum 3.58 wt% Silicon 14.62 wt% Phosphorus 0.30 wt% Chlorine 0.11 wt% Potassium 0.16 wt% Calcium 0.16 wt% Iron 1.64 wt% K Ca Fe -4-Fe keV F igu re 3.7: Rubber Gasket Composition next section) from the blank experiment, with a range of shapes and sizes. It can be seen that the impactor is effective at its designed cutoff size of 2 /xm. The sizes of Lexan fragments can also be very small and approach those of the soot chains, resulting in the impactor improvements discussed above. These sizes of Lexan contamination will be very difficult to separate by inertial effects, and will likely require detailed analysis of the various Aethalometer beam signals to determine its levels (i.e. search for possible Lexan signature). Table C.4 shows a higher particle density than the blank run, with some soot particles present. This grid sample also shows numerous small chain-like particles stuck on the laces with comparable darkness, indicating similar light element compositions (e.g. C, O, H) as the laces. However, the confirmation of soot particles (using E D X analysis) will be performed on the T E M , due to resolution limitations of the S E M . The absence of dust particles can be due to their complete burnout in the presence of a flame, as can be the case for other contamination sources in Table C.2. 3.4. Contamination Control 43 F i g u r e 3.8: Stainless Steel Diaphragm Particle T E M Ana l ys i s To confirm Lexan and soot particles identified with the SEM, the lacy carbon grids are re-analyzed with the T E M , which allows up to 200,000 times magnification. This level of detail also enables soot particle structures from shock tube combustion to be analyzed and compared to conventional chain-like aggregates [26]. Additional information from the sample grid such as particle size, density, and capture efficiency can also be estimated to eventually validate the Aethalometer black carbon mass measurements. Shock tube exhaust gas (from blank and sooty experiments) is passed through the grid using the sampling setup shown in Figure C.16. For each sample, a flow rate of approximately 100 mL/min is used for 10 minutes, controlled by a portable gas-sampling pump. The T E M grid is held in place by | " tubing against the lip of a Swagelok connector, with the pump drawing a small portion of the main flow. Figure C.17 shows an enlarged view of the background lace netting strung between the main grid boundaries, which are used to trap the particles in the flow. A representative grid square is then used to search for particles attached to the laces. Due to the higher resolutions of the T E M , neighbouring and background element contamination is more prominent, and complicated beam correction factors are required to obtain accurate ele-mental percentages. Therefore the particle composition chart will always contain carbon (from the laces) and copper (from the grid boundaries), with any additional elements 3.4. Contamination Control 44 kCts E l e m en t C o n c e n t r a t i o n Carbon 12.01 wt% Oxygen 38.16 wt% Sodium 0.86 wt% Magnesium 0.34 wt% Alum inum 0.48 wt% Silicon 21.05 wt% Sulfur 0.27 wt% Chlorine 0.12 wt% Calcium 0.06 wt% Chromium 4.61 wt% Manganese 0.46 wt% Iron 19.37 wt% Nickel 2.20 wt% 0 2 4 6 8 keV F igu re 3.9: Stainless Steel Composition used to help identify the material. From the blank experiment grid, Figures 3.10, C.18, C.19, and C.20 show various Lexan diaphragm particles (rectangular shaped) stuck on the laces, with their composition shown in Figure C.21. This confirms the above S E M results, where soot-sized Lexan particles exist in the gas sample and need to be removed or accounted for in the measured black carbon mass. In addition to the diaphragm parti-cles, the sooty experiment grid also shows typical soot-like aggregates stuck to the laces (Figures 3.11, C.22, and C.23), with the E D X analysis confirming the main carbon com-ponent without the silicon peak (Figure C.24). Detailed T E M mode photographs of some representative soot particles are shown in Figures C.25 and C.26, taken at the highest magnification (1 mm = 5 nm). These images show the typical chain-like structure of conventional soot particles, consisting of individual primary spherical particles in the 10-50 nm diameter range and total aggregate sizes of 300-500 nm. Finally, representa-tive particles analyzed from both grids are summarized in Tables C.5 and C.6. It can be seen that individual and short chains of primary soot particles are also present on the blank experiment grid, indicating possible wall deposit and/or sample line contamina-tion discussed above. However, the sooty experiment grid shows larger quantities of long conventional soot chains, which could be an indicator of freshly created soot particles in 3.5. Particle Loss Minimization 45 the flame rather than soot contamination. These long chains also present a challenge to future impactor designs, where the random chain alignment in the flow can greatly affect the target impaction/aerodynamic diameter. F igu re 3.10: Lexan Particle (1 /xm) 3.5 Particle Loss Minimization While contamination control was primarily concerned with blank experiments and reduc-ing their apparent black carbon results to 'clean' gas levels, particle loss minimization mainly involves soot-producing experiments and semi-theoretical calculations to study various soot particle loss mechanisms pertaining to the shock tube apparatus. As in the contamination control case, segments in the overall sampling system are isolated and ana-lyzed in terms of applicable aerosol loss phenomena, with subsequent control steps aimed at eliminating or reducing their effects on the desired measurement quantities. In most cases, the goal is to prevent soot particles from adhering to wall surfaces along its path of travel, in order to collect the maximum fraction of soot produced from combustion on the Aethalometer measurement tape spot. Appropriate Aethalometer data analysis procedures (shown later) will be used to account for losses that cannot be completely 3 . 5 . Particle Loss Minimization 46 mm Figure 3.11: Soot Particle eliminated, to ensure that the measurements accurately represent the P M produced from the combustion event. 3.5.1 Ins ide Shock Tube While the combustion product gases are still in the shock tube, the main loss mechanisms to be considered are thermophoretic, gravitational settling, and diffusional transport. Thermophoretic losses take place in the short interval between the times of combustion and when the gases reach ambient temperatures. Settling and diffusion losses occur during the settling time used to filter out diaphragm particles. In both cases, any contact with the shock tube walls causes the van del Waals adhesive forces to become sufficiently strong to cause permanent attachment and loss of the particle. These forces are based on the attraction between dipoles caused by random movement of electrons in both materials. The net adhesive force between a particle and a plane surface can be expressed as F a d h = 12S ^ 3 '^ Calculations using typical shock tube parameters are shown in Appendix D , where the 3.5. Particle Loss Minimization 47 adhesive forces can become extremely large. In addition, the separation distance contin-ually decreases (while increasing the van del Waals force) after initial contact, until the attractive forces balance particle deformation forces. Since adhesive forces on micron-sized particles can exceed other common forces by orders of magnitude [16], surface contact resulting from particle migration toward the walls should be avoided wherever possible. Thermophores is Due to the large temperature gradients established behind the reflected shock as well as within regions of combustion, thermophoresis effects will accelerate the rate of particle motion toward the colder shock tube wall surface, and thus increase the likelihood of adhesive losses. For small particles (d < A), the thermophoresis force arises from a greater transfer of momentum from the surrounding gas molecules on the hot side of the particle relative to those on the cold side [16], as a result of the temperature gradient. Due to the high mobility of small aerosol particles and the existence of significant temperature spikes during the combustion event, large thermophoretic forces and velocities can occur. It can be seen that the velocity is independent of particle size and directly proportional to the temperature gradient (assuming negligible temperature gradient within the particle). Appendix D shows calculations of typical Fth and V t / , under shock tube conditions. Since the tube diameter is relatively small, particle velocities in the range of 0.4 cm/s will only take a few seconds to reach the tube walls, depending on the temperature gradient distribution. Therefore, it is necessary to withdraw the combustion gases as rapidly as possible immediately following the experiment (but without disrupting the combustion process) to minimize this form of loss. -p\d?VT f (3.2) VtH -0.55r?Vr PgT (3.3) Se t t l i ng Since gravitational settling is required to remove contamination particles in the form of diaphragm fragments, possible soot particle losses during the settling period also need to be considered. For a representative soot particle in the quiescent post-experiment gas 3.5. Particle Loss Minimization 48 environment, the settling velocity is a balance between drag and gravity forces. Assuming no slip factor correction ('slip' at the surface causes sub-micron particles to settle faster) or equivalent aerodynamic diameter correction, the terminal settling velocity can be expressed as _ pPd2g V V S - — (3.4) Sample calculations of settling velocities are shown in Appendix D, where it can be seen that the velocity increases rapidly with particle size. Typical soot particle size ranges result in 2 x l 0 ~ 5 to 7 x l 0 - 3 cm/s. Therefore, using a conservative estimate of the starting position in the middle of the tube, it will take between 7 and 2500 minutes to lose the soot particles due to settling, depending on the size. Without any knowledge of the expected soot particle size distributions, the time required to settle out sufficient diaphragm contamination is used as the settling period, in an effort to minimize settling losses. D i f fus ion During the post-experiment settling period, soot particles will also experience diffusion which will result in increased losses to the shock tube walls. Analogous to gas molecule dif-fusion, particle transport occurs in a concentration gradient in the direction of decreasing particle concentration. Since the soot particles are initially produced in the flame (near the centreline of the tube), the particle concentration gradient extends radially toward the walls. The associated particle diffusion coefficient represents the speed of its trans-port, which is related to the intensity of its Brownian motion (random wiggling motion caused by surrounding gas molecules). For sub-micron particles, diffusional transport is important since the diffusion coefficient is inversely proportional to the particle diameter, as shown below. B = i § '• < 3 5> The flux of soot particles is directly proportional to the diffusion coefficient, with sam-ple calculations using typical sootl-sizes shown in Appendix D. It can be seen that the number of particles diffusing through a unit cross-section per second in a unit concentra-tion gradient increases rapidly with decreasing size. As a result, smaller (primary) soot 3.5. Particle Loss Minimization 49 particles will be lost first during settling, followed by progressively larger chain aggre-gates. In addition, sub-micron particles with non-negligible slip factors will diffuse even faster with decreasing size (D proportional to d~ 2). For example, a 50 nm particle will be transported approximately 40 times faster than a 500 nm particle under the same conditions. Therefore, the settling procedure used to filter out diaphragm contamination will also cause diffusional losses of small soot particles. Although the current procedure is to minimize the settling time in order to minimize this form of loss to the tube walls, further studies are needed to find the acceptable compromise between the two processes. The relatively small-sized particles lost will likely not represent a significant portion of the total black carbon mass produced. 3.5.2 Gas Venting Process After the settling period, the shock tube contents are released through the valve at the end of the driven section (see Figure C . l l ) . To minimize particle losses during this venting process, appropriate valve and tubing selections must be considered. At the same time, the level of mixing of combustion gases, dilution gas, and particulate matter should also be considered in this context, as a uniform gas sample composition greatly simplifies subsequent P M sampling and analysis (see Instrument Detection section). Since combustion takes place in the experimental region, the valve should be placed on the exhaust port that is closest to the end of the driven section. This ensures maximal P M extraction by minimizing the distance and time for particle transport losses (due to fluid motion) in the vicinity of shock tube exit opening. High venting flow rates are desirable to minimize thermophoretic losses inside and downstream of the valve, where the rapid gas expansion process significantly cools the physical metal components. Reducing particle residence times in this radial temperature gradient (toward the valve and tubing surfaces) will minimize thermal losses. In addition, larger flow rates will increase turbulence in the gas /PM flow downstream into the sample container. The combined effects of molecular and turbulent diffusion enhance mixing of the gases and particles, resulting in more uniform sample container contents. Due to pre-fabricated shock tube port opening (~ 4.3 mm ID) restrictions, the flow rate through the valve is maximized using a | " ball valve with a fully open inner diameter of 4.5 mm. Trials of venting helium into atmospheric pressure using various simulated post-experiment 3.5. Particle Loss Minimization 50 pressures are shown in Table D . l , where the total shock tube contents from 20 to 40 bar experiments (i.e. post-experiment pressures of ~ 14 and 22 bars) take a reasonable amount of time to completely evacuate. Since the entire tube contents need to be drawn into the sample container to achieve proper mixing, minimizing the total durations will minimize the level of thermophoretic and transport losses during the venting process. In addition to flow rate considerations, the tubing from the valve to the sample container (via the particle impactor) should also minimize particle transport losses. For example, it needs to have internal diameters larger than the shock tube exit (i.e. minimum | " thin-walled) to avoid further flow restrictions. The connection should also have minimal, length and obstructions to reduce adhesion and thermal losses, as well as being electrically conductive to reduce electrostatic losses. As a result, a short straight section of | " stainless steel tubing is used up to the 1" impactor inlet, as seen in Figure C . l l . 3.5.3 Samp le Con ta ine r A n appropriate g a s / P M sample container is needed between the shock tube and the particle sampling instrumentation, for several reasons. The container should be able to achieve lower particle losses compared to the shock tube, provide a uniform gas /PM composition for sampling, and provide an atmospheric downstream pressure to maximize the above-mentioned venting flow rates. The container also needs to be durable and easily reusable for various test conditions. Since bag sampling is sometimes used as an intermediate reservoir in engine emissions tests, a similar approach is attempted using a large flexible bag (fabricated for this specific purpose) as the sample container. Deve lopment Due to the large amounts of dilution gas, typical 40-bar experiments produce up to 450 liters of gases at atmospheric pressures. The required dimensions (40"x90", 550 L capacity) are obtained by extrapolating from the volumes and dimensions of standard bag sizes. Smaller sizes can be made for low pressure experiments, in order to keep a favourable volume-to-surface area ratio. The final product is constructed by cutting open smaller bags, and heat-sealing to make the required size. A custom seal with foam mounting tape and metal washers is used in conjunction with a bulkhead fitting to make the bag opening. Contamination tests are performed with filtered helium filled into the bag with 3.5. Particle Loss Minimization 51 subsequent P M sampling, while leakage is checked by settling the filled bag contents. It was found that periodic flushing of the bag is needed to reach clean (atmospheric) background levels shown in Table 3.1, while leakage during a typical sampling duration is less than 5%. To further increase turbulence and mixing in the bag, slight flow restrictions in the form of an elbow can be placed at the end of the bulkhead fitting inside the bag entrance, acting as a nozzle. This also serves to direct the main flow down the length of ,the bag (away from the surface), thus reducing particle impaction losses. Further mixing prior to sampling can be achieved by manual bag agitation/shaking. In terms of material properties, the bag should be sufficiently thick to prevent permeation, while the inner surface should be inert to minimize absorption/adsorption of gases and compounds. In addition, the material should also be opaque to prevent reactions of gas species (e.g. N O x ) with ambient light. As a result, two different materials with the above properties are tested for their suitability, as shown in the next section. Losses The particle loss mechanisms of concern while the gases are in the bag include diffusion, gravitational settling, and electrostatic losses. Diffusion and settling of soot particles (via the same mechanisms mentioned above) are minimized by the large volume-to-surface area ratio and distances to bag surfaces. Electrostatic losses, however, can be significant, for charged particles in the vicinity of non-perfectly conducting material surfaces. For particles formed in the flame (e.g. soot), a net charge is acquired through direct ioniza-tion of surrounding gas molecules in the presence of high temperatures [16]. The resultant electrostatic force (using Coulomb's law) and the associated terminal velocity (using the Stokes drag relation) expressions are shown in Appendix D. Since the electrostatic force on highly charged particles can be several orders of magnitude larger than gravitational forces, the sample bag material should be highly surface conductive (in addition to being volume conductive) to minimize the associated losses. The difference in the Aethalometer results due to electrostatic deposition can be seen in Figure 3.12, where a non-conductive (polyethylene) surface is initially used, before covering the entire inner surface with highly conductive aluminum foil. The results (using similar initial black carbon quantities) are compared to the bag made from a fairly conductive carbon impregnated polyolefin ma-terial. In the non-conductive case, particle losses observed from the Aethalometer data can be as high as 2% per minute for sooty experiments. The observed particle losses 3.5. Particle Loss Minimization 52 with the highly conductive aluminum surface are likely due to settling and diffusion, especially with the decreasing volume-surface area ratio. The additional observed decay in the less conductive polyolefin surface are likely due to some electrostatic losses of soot particles as well as settling of larger contamination particles. Since the decay in measured soot quantities can be accounted for in an appropriate Aethalometer data analysis, the conductive polyolefin bag material will be used due to its cost and ease of manufacture. 10 20 Time (minutes) 30 40 •Non-Conductive Conductive (Al layer) Conductive (Surface) F i g u r e 3.12: Electrostatic Loss Comparison for Various Bag Materials 3.5.4 Container to Aethalometer Particle losses in the path from the sample container to the sampling instrument are mainly due to transport and electrostatic mechanisms. Since the flow velocity is much smaller than the valve venting process, the conductive tubing used should minimize these losses. Transport losses in the forms of adhesion, diffusion, and settling are typically reduced by using short, straight tubing with large flow areas. However, since the Aethalometer inlet flow rate is constrained by external considerations, particle residence times (and losses) will increase with increasing tube diameters. In order to keep a relatively constant flow area prior to the | " Aethalometer sampling chamber, | " diameter aerosol sampling tubing wil l be used as a compromise between particle-3.5. Particle Loss Minimization 53 to-surface distances (~ 3.9 mm from centerline) and residence times (~ 1.7 seconds, based on 4 L P M flow rate and 2.5 m length). Electrostatic losses are minimized by the surface conductive silicone material in the tubing, which is commonly used in aerosol experiments where static charge buildup is reduced in the same manner as the conductive bag surface. The importance of electrically conductive tubing can also be seen in Figure 3.13, where experiments are performed to investigate the extent of particle line losses due to electrostatics. By alternating the sample flow between conductive and non-conductive (plastic) tubing of similar lengths, it can be seen that electrostatic losses will greatly affect the desired amount of black carbon particles measured by the Aethalometer. In addition,.Figure 3.14 (using a similar line loss setup) shows that minimizing the total sample line length is not critically important to control various transport phenomena, as the particle residence times are short enough to avoid significant losses. 45 40 E" 35 E 2 3 0 I 25 | 20 o in £ 10 C - conductive, N - non-conductive •< > c < 3> N — : « . •< s-0 «s > N <=—=> N • <= 5»-c • — • 10 15 20 25 Time (minutes) 30 35 40 F i g u r e 3.13: Conductive vs Non-Conductive Line Loss Experiment Based on previous thermphoretic loss considerations of sub-micron sized particles in tube flows [48], the deposition efficiency due to diffusion and electrostatics are compa-rable to those from thermophoresis. The significant electrostatic effects observed also agreed with the findings in [48], where particles in charge equilibrium were shown to have non-negligible electrostatic deposition efficiencies. As a result, further experimental 3.6. Instrument Detection 54 50 45 g- 40 1,35 c r 30 c E 25 o 20 8 15 to 5 10 5 0 - A L - long, S - short •«? => • «s =>| L S «s > • • • • L • • < => • S < - > i • c • e => « — o • L • • — * i i — — 1 1 —i 1 _ 10 15 20 25 Time (minutes) 30 35 40 F igu re 3.14: Long vs Short Line Loss Experiment and numerical work on quantifying soot losses based on each relevant mechanism should be undertaken. 3.6 Instrument Detection While the previous methodologies improve the experiment quality and the amount of measurable particles of interest, the actual sampling instrument and data interpreta-tion used for these measurements are just as vital to the overall objectives. Due to the relatively small amounts of fuel present, it becomes a challenge to accurately and con-sistently measure the levels of resultant particulate matter from the combustion event. Typical non-premixed methane outputs of 200-1000 ng roughly correspond to the mass of one equivalent 25-45 /mi radius carbon particle, thus extremely sensitive and dedi-cated instrumentation is required. Conventional direct mass instruments (e.g. T E O M ) require relatively large and steady P M mass flows such as those from steady engine operation. Gravimetric analysis of filter samples weighed before and after particle col-lection also requires at least 10 mg to obtain a meaningful signal. However, the most important drawback of these instruments is the inability to separate various particle species based on their unique properties. As mentioned in the literature review, previous 3.6. Instrument Detection 55 shock tube P M experiments produced sufficient amounts to be measured by direct meth-ods, while previous Aethalometer applications were aimed at ambient environment black carbon monitoring. For the present experimental objectives of obtaining the amount of black carbon from each shock tube combustion event, appropriate adjustments to the Aethalometer measurement and interpretation procedures must be made due to various experimental and instrumental constraints. 3.6.1 Operation Principle To enable the real-time global measurement of extremely small amounts of black carbon species in the PM/gas sample stream, an instrument based on the optical absorption principle must be used. The black carbon component in atmospheric aerosol samples (from the combustion of carbonaceous fuels) can be defined chemically and specifically [14] (i.e. the fraction of aerosol P M that is insoluble in polar and non-polar solvents; stable in a pure oxygen atmosphere to 350°C; displays the Raman spectral lines characteristic of both the graphite structure and the features of microcrystallinity, etc.). The Aethalometer principle is based on the operational definition of strong optical absorption of the B C species in the visible spectrum. The actual calibration of the optical absorption measurement is performed by a chemical analysis of the CO2 produced by combustion of a sample after extraction and thermal pre-treatment, where the mass yield of carbon (in /jg) is compared to the sample 'blackness' in terms of the known B C mass. The 'blackness' measurement was found to be sensitive Only to the amount of carbon thus defined (i.e. insensitive to any other aerosol species that contribute to the total particulate mass) [14]. Therefore the level of visible light absorption can be interpreted directly to the mass of B C , since it is the only optically absorbing species in the spectrum. More specifically, the Aethalometer measures the attenuation of beams of high-intensity light transmitted through a quartz fiber filter that is continuously collecting the inlet gas sample. The light shines through the aerosol deposit spot, penetrates the diffuse mat of filter fibers, and is detected by a photodiode directly underneath the filter support mesh grid. A reference beam signal is measured by another photodiode under a clean/blank portion of the tape (adjacent to the particle-laden spot), which corrects for fluctuations in the beam intensity of the common incident light. By using the appropriate value of the specific attenuation coefficient (see Appendix E) for the particular filter and optical com-3.6. Instrument Detection 56 ponent combination, the black carbon content of the aerosol deposit can be determined from the measured attenuation at successive time intervals. As the black carbon content of the aerosol spot increases, the photodiodes detect diminishing amounts of light (i.e. increasing attenuation), which is linearly proportional to the black carbon mass in the gas sample stream. Numerically, the optical attenuation (ATN) and specific attenuation (a) are defined as (details in Appendix E) r j ATN = -100 x ln( T ~ ) (3.6) Io — do AATN . . a = (3.7) A M , BC (Flow Rate) x (Time) ^3'8) Although the amount of ambient light passing through the filter, support mesh, and gas is small, a correction for this dark response signal of the system is required for accuracy. Optical measurements are therefore first taken with the light source turned off, to detect the electronics' zero signals in both the sensing and reference beams (i.e. d and d 0 , respectively). After the system stablizes with the light source turned on, the transmitted light intensities from the sensing and reference beams (I and I0, respectively) are measured. The factor of 100 is used for numerical convenience. By taking into account the zero offsets from each measurement cycle (i.e. correcting for possible variations in light intensity outputs), the resultant logarithmic ratio represents the true detector responses to the incident light beams. The starting attenuation value (ATNo) is then found at the beginning of the measurement cycle, using Equation 3.6. At the end of each timebase period, the new attenuation value is calculated, with the increment due to the amount of black carbon deposited on the filter spot. The difference in the A T N values is directly proportional (subject to the relevant assumptions) to the black carbon surface loading, which can be converted to mass by using the collection spot area. The relevant B C concentration during the corresponding timebase period is then obtained by dividing the mass by the total volume of inlet gas, based on the sensed flow rate and time. In addition to the infrared beam channel used for black carbon, the ultra-violet beam channel gives an increment of BC-equivalent mass of UV-absorbing material (e.g. fresh diesel exhaust). In theory, if all particles in the gas sample were broad-spectrum 3.6. Instrument Detection 57 absorbing (i.e. characteristic of black carbon), then the two channels should give the same mass results. Therefore, their difference can be used to determine the amount of spectrally-specific absorbing particles (e.g. possible signature for white Lexan particles) in the total P M mass, and therefore aid in the detection of contamination levels. The appropriate a for each measurement wavelength (Table E. l ) are empirically determined by comparisons with other available measurement techniques. Sample calculations using typical B C data are shown in Appendix E, with the same method applicable to the U V channel results. The Aethalometer is commonly used to measure small amounts of particulate species (especially black carbon) in various ambient environments [4] [10] [34]. At extremely low B C concentrations such as those from the shock tube exhaust gas, the effective noise levels should be very small in order to achieve an acceptable signal to noise ratio for proper data interpretation. The Aethalometer detectors and electronics can usually resolve an increment of less than 1 ng of B C on the filter. This sensitivity translates to output concentrations of 250 ng /m 3 at typical settings of 4 L P M flow rates and 1-minute timebase periods. The high sensitivity is critical for the shock tube application as blank experiment soot levels need to be accurately measured and used as baseline results. The continuous optical analysis with real-time output of results is also important in the data analysis algorithm, as the minute-to-minute black carbon amounts are valuable information to the determination of contamination and desired quantities. In addition, the Aethalometer is conveniently self-contained, automatic, and uses no consumable materials. Furthermore, the specific optical attenuation coefficients are empirically determined for this particular instrument and measurement principle. These values are compared extensively with other analytical techniques, with favourably results as outlined in Appendix E [14]. 3.6.2 Data Algorithm Modifications Due to several instrument and system limitations, the black carbon data interpretation algorithm in the Aethalometer needs to be modified to suit this shock tube application. The instrument is highly sensitive to inlet condition fluctuations (e.g. pressure, gas composition, relative humidity, etc.), and is designed to operate at constant atmospheric conditions. A few minutes of stablization time are typically required when new inlet conditions are introduced. For example, the change from sampling ambient room air 3.6. Instrument Detection 58 to the shock tube gases causes a different pressure (due to the lighter helium) to be exerted on the filter tape fibers, and the resultant fiber compression relaxation causes the equilibration period. The light beam attenuation values will also change significantly when passing through the predominant helium gas, which combine with optical deviations to cause large offsetting spikes in unprocessed Aethalometer data (Figure 3.15). In addition, the internal thermal mass flow meter will respond to inlet gas property fluctuations by keeping a constant mass flow rate (with the corresponding volume flow calculated from the calibration gas properties), as described in Appendix F. Due to the large difference in helium and air properties, the apparent sensed volume flow rate must be corrected to determine the true volume of shock tube sample gas through the Aethalometer. F i gu re 3.15: Data Spikes Due to Gas Composition Changes To overcome these instrumentation issues, the experimental and data analysis procedure are slightly modified. The large sample bag collection method is ideal for providing a uniform gas composition and constant atmospheric inlet pressure. This serves to minimize the stablization period by limiting the data spikes to the initial and final gas switches. By maximizing the amount of valid data in Figure 3.15, detailed analysis such as possible particle losses and background contamination can be used to determine 3.6. Instrument Detection 59 the true black carbon mass in the gas sample. The deviations in the optics when the Aethalometer inlet is subjected to the helium-based sample gas, however, cannot be completely avoided. As seen in Table E.3, the optical attenuation value decreases drastically when the inlet gas changes from air to helium, with the opposite effect at the end of the bag sampling duration. The existing data algorithm automatically interprets these changes as,large negative and positive black carbon amounts, which are physically incorrect. Since the optical configurations and filter fibers rearrange after each 'shock' to the system and do not adjust back to their exact original conditions, the corresponding offsets in A T N are not necessarily equal. Therefore only the data during the shock tube gas (i.e. bag) sampling period is used in the analysis to eliminate this effect. In addition, the accurate volumetric flow rates must be known, especially to combine with the manual mass integration procedures (described below) to find the total mass output. Appendix F gives a detailed account of the Aethalometer's mass flow meter calibration, using an external reference volumetric flow meter. Interpolation over the ranges of the two variables (helium volume fraction and flow rate setting) is used to obtain the true volume flow rate of all gas samples through the Aethalometer. Furthermore, the existing data algorithm gives outputs in terms of black carbon mass concentration, which is calculated from the mass increment and the gas sample volume during the timebase period. Since the internal flow meter is relatively inaccurate [10], correcting the minute-to-minute mass concentrations would be required, in addition to the external flow calibration. Rather than converting the concentrations back to mass, it is more convenient to directly use the intermediate A T N values to manually integrate the minute-to-minute B C mass increments, as shown in Appendix E. By using the optical attenuations in the data file directly, the total mass will be independent of flow rates, gas composition changes, and temporal fluctuations. The revised, Aethalometer data processing algorithm and sample calculations are summarized in Appendix E, using the sample B C channel data in Table E.2 (UV channel data is modified in a similar manner). It can be seen from Table E.4 that only the A T N column is used to find the minute-by-minute B C mass increments directly for the manual integration method, using the specific attenuation coefficient (a) and the collection spot area. The internally calculated concentrations are used as a reference cutoff point, where the first positive value is used as the first minute in the table to 3.6. Instrument Detection 60 avoid negative results. The last entry in the table is taken as the last concentration that appears to be in the range of the valid data, before the sudden increase due to the gas composition change. The black carbon mass increments are then found from Equation E.13, starting with the difference between the first and second ATN values. Figure 3.16 is then plotted using Table E.4, giving a minute-to-minute mass deposition rate as a function of sampling time. The minute-by-minute mass increment progress is a very important observation in this data analysis procedure since discretion must be used in each individual case to decide where the valid data range begins and ends. The initial increase in the curve is due to the instrument stablization period, where the disturbance of helium causes fictitious attenuation values. The system subsequently adjusts to the new uniform inlet gas composition, resulting in the mass increments settling down to a steady value before any significant losses take place. After the settling period, the curve will most likely decay due to progressively higher particle losses (e.g. increased settling and diffusion as the bag's volume to surface area ratio decreases). In order to accurately calculate the total black carbon mass produced from the experiment, the most representative mass deposition rate is determined using all the valid data points. By taking into account all relevant data in the sampling duration, an appropriate empirical trendline can be fitted with greater confidence, as well as locating possible external effects such as contamination and particle losses. Experimental investigations with periodic sampling of the same bag contents (e.g. Table E.5) show that particle losses typically exhibit exponential decay characteristics. Assuming no other significant contamination effects (e.g. particle entrainment from bag surfaces), the initial black carbon concentration gradually decays over time, as shown in Figure E . l . Appendix E also shows trendlines for other common mass deposition curve shapes encountered. Since exponential decay is prevalent in most experiments, it will be used as the default curve fit. More detailed knowledge of the types and rates of particle loss mechanisms in the sampling process are needed to determine a better correlation of the data points, and thus a more accurate total BC mass. Finally, this curve-fitted initial mass deposition rate is multiplied by the amount of time needed to sample the entire bag contents to arrive at the total black carbon mass resulting from the shock tube experiment. Table E.6 and Figure 3.17 give an example of the UV channel application described above, where the two time-dependent decay curves are superimposed on the same graph. It can be seen that the U V P M quantity is consistently higher than the BC species, which 3.6. Instrument Detection 61 12 - 1 0 E 1 8 E 6 s> o 4 in in • " • • • • • • • • 10 15 20 Time (minutes) 25 30 Figure 3.16: Typical Black Carbon Sampling Results is expected from the U V wavelength absorption of various contamination particles. The U V channel data also displays a fairly constant decay, which could be caused by grav-itational settling and transport losses of large diaphragm particles. The consistent B C values during the sampling period suggests that the soot particles in the bag (at this initial concentration range) are not systematically lost via typical mechanisms such as electrostatics and thermophoresis. Table E.7 and Figure E.6 show an example of a high soot-producing experiment under premixed methane/air conditions, where the particle decay is more prominent due to higher initial concentrations. It can be seen that signifi-cant soot losses to the bag surface can occur if the initial concentration exceeds a critical value. It is also important to start sampling each sooty experiment on anew tape spot (as well as using the first few minutes of valid data) to eliminate the apparent attenuation decrease due to scattering effects [14]. Furthermore, as shown in Table E.2, the reference beam signal voltages fluctuate throughout the sampling period and is recorded in the instrument log file, where a few hundred ppm is deemed acceptable [14]. 3.6.3 Error and Uncertainty Since diaphragm particle contamination is still possible after settling and impaction procedures, the U V channel data is also used to determine the level of contamination 3.6. Instrument Detection 62 A B C >UV * • • • , — : A — : A A A A A A A A A A A A A A A * A A A A A A A A A A A A A A A A A A A A A A A A A A A A * A • A I ; | 1 . 1 0 10 20 30 40 50 Time (minutes) Figure 3.17: Typical Black Carbon and U V P M Sampling Results in the B C channel signal. As explained in the example in Appendix E, the current approach is to compare the real-time mass increments from the two different wave-lengths (Figure 3.17), over the entire sample duration. The additional mass from the U V light absorption is likely due to the non-BC material on the tape filter spot. For example, white Lexan (compared to dark steel shim diaphragm) particles have more spectrally selective absorption characteristics, resulting in an increased discrepancy between the two mass increment curves with increasing Lexan contamination. The shapes of the two curves also provide certain clues to the types and amounts of contamination particles. The relatively larger Lexan particles (see S E M / T E M results) settle at a faster rate in the sample bag, resulting in a larger decay in the U V mass curve. Therefore, the amount of deviation between the two curves can be used to esti-mate the level of larger particle contamination in the B C signal over the sampling period. Quantitatively, however, the U V P M mass data can only be used to estimate the amount of contamination. Due to the non-physical definition of the U V P M material as a BC-equivalent mass (see Appendix E) , the amount of actual material absorbing in the U V wavelength cannot be directly calculated. Further studies on the absorption charac-teristics of different types of contamination particles under U V light are needed, as well •c* 20 E ~5) c r 15 c a> E a> b 10 _c in in «5 r-3.6. Instrument Detection 63 as its relation to the BC-equivalent mass definition. In addition, white particles such as Lexan fragments can be charred/darkened in the high-temperature experimental region, resulting in increased broad-spectrum visible light absorption. This further complicates the attribution of sources of increased apparent B C mass signals. Therefore, comparisons using Figure 3.17 can be only used as a qualitative method at this point, instead of a di-rect subtraction between the two mass quantities. Further studies are needed to arrive at an acceptable accounting procedure for the amounts of all existing contamination species. In addition, the Aethalometer's general optical measurement technique has various docu-mented uncertainties. First of all, the optical signal variability can be significant for some units tested [10], making the external factory optical calibration questionable. Another common effect is decreased optical path with increasing particle loading on the filter tape [52]. This is attributed to the 'shadowing' of the particles in the deep fiber matrix, which is more pronounced for fresh soot particles. Although empirical corrections are possible, this effect is minimized by starting every experiment with a clean tape spot and using a smaller maximum attenuation value (before tape advance occurs). The non-uniform distribution of foreign contamination in the path of the beams will not allow the correct compensation of reference beam fluctuations. Most importantly, however, it is found that the specific attenuation coefficient (o~) can be dependent on factors such as black carbon mass fraction and type of atmospheric environment. The a from the man-ufacturer calibration is a combination of absorption and scattering contributions [40], where the variations in the scattering a component can result in overall a values be-tween 8 m 2 / g and 19 m 2 / g . Based on the testing location conditions, the difference in overall a is a result of the aerosol mixture characteristics such as black carbon content, production process, and aging phase [35]. Due to the fractal nature of B C particles, no size dependence on a is observed. Since the a value of 16.6 m 2 / g is obtained from an external calibration using a test sample under ideal conditions, the determination of the applicable a for the shock tube aerosol and gas mixture should be performed. Exper-imental methods [17] [32] to calibrate the Aethalometer's optical coefficients for shock tube relevant P M samples will allow more accurate quantitative interpretations of the black carbon channel signals. 3.7. Conclusions 64 3.7 Conclusions Due to the small amounts of particulate matter expected from the shock tube and methane flames, the development of an appropriate experimental methodology is necessary to achieve the overall sampling system objectives. The uniqueness of the shock tube facility, along with the novel measurement concept, posed significant technical challenges. Although the effects of various critical experimental issues are often related and approached in parallel, systematic progress have been made in each category, including suggestions for further improvements. Experiments based on methane flames should be performed to validate and justify the practically of this P M sampling system and methodology. Diaphragms made from standard thickness shim stocks provide better bursting con-sistency, price, and pressure ranges than previously machined pieces. However, local pressure disturbances, high throttling losses, and black carbon signal contamination are created by the dislodged fragments, resulting in high experimental condition variability. Custom traps and magnetic filtering were generally ineffective. The use of Lexan di-aphragms eliminates these disturbances and provides favourable bursting and throttling loss characteristics, thus the ability to achieve target experimental temperatures and pressures. Diaphragm-free systems such as a rapid gate opening can eliminate this form of contamination. Careful control procedures are needed to prevent foreign particle contamination, enable the accurate measurement of soot particles, as well as to prevent disturbances to the combustion and soot formation processes. Target baseline B C levels are around 150-200 ng, based on normalized sample volumes. Diagnostic tools (e.g. B C signals, light emis-sions, camera frames, electron microscopy) applied to various control experiments (e.g. non-shock, blank, and combustion) confirm the existence of a variety of contamination species and sources. Microscopy results also show that contamination particles generally differ in size and structure, which can be used to aid the design of control procedures and apparatus. Vigorous cleaning and entry prevention efforts do not consistently reduce the initial high blank test levels, which is also true for some post-experimental control techniques. Particle impaction and settling (although somewhat qualitative) is the most effective method to minimize B C signal contamination. However, further impactor 3.7. Conclusions 65 design and settling duration studies are needed to achieve the desired cutoff size with high efficiencies without filtering out soot particles, using additional particle size and structure distribution information. Improved tube surface finish techniques will also improve the cleaning efforts of residue particles. Soot particles are mainly lost through thermopheoresis, settling, diffusion, and electro-statics, within each segment of its path from the flame to the Aethalometer. Although techniques such as maximizing flow rates and minimizing residence times are employed to reduce losses, fully quantitative accounts require detailed size distributions. A flexible surface conductive sample bag is developed to enhance PM/gas mixing, reduce losses, and to provide an intermediate reservoir to meet Aethalometer constraints. Further reductions in particle losses can be achieved by a large custom shock tube port opening, in conjunction with a fast acting gas-sampling valve. Further studies on the compromise between diffusion soot losses and contamination particle settling should be performed, as well as improved sample bag designs. Surface conductive material should always be used to minimize significant electrostatic forces. Aethalometer black carbon data analysis algorithms were modified to take advantage of the attenuation signals directly. The total B C mass from the experiment is determined from the most representative mass increment (during the first few minutes of the time-dependent decay curve) and the time needed to sample the entire volume. Typical B C and U V P M mass increment curves decay due to particle losses and optical effects. Although the U V channel can be used to obtain the level of contamination and correct for the apparent B C mass, the method's quantitative validity requires further study. In addition to particle loss mechanisms, sooty premixed experiments have been useful to the study of Aethalometer signal decay and corresponding data analysis. External validation techniques to find the most relevant specific attenuation coefficients should also be performed. The combination of pre-experiment and post-experiment control procedures have reduced blank test baseline levels to 100-150 ng. The remaining background B C is likely due to a combination of sources, and requires further studies to minimize or eliminate their exis-tence. It is extremely difficult to achieve the desired pristine experimental environment, due to the effectiveness of shock waves in particle entrainment. Although contamination 3.7. Conclusions 66 and particle loss considerations are somewhat qualitative, the available Aethalometer data can be reliably used to determine the black carbon mass produced from each ex-periment, provided that it is sufficiently above the baseline levels. The premixed and non-premixed methane control experiments have been helpful to locate the critical ex-perimental challenges, and significant progress has been made on each issue. Chapter 4 Particulate Matter Sampling from Methane Flames 4.1 Introduction While research efforts in diesel-fuel alternatives usually focus on ignition and combustion feasibility, gaseous and particulate emissions characteristics are equally important, as soot and N 0 X are notorious diesel engine pollutants. However, the combustion variable effects on P M emissions are difficult to encapsulate in engine testing or numerical simulations, due to the complex interactions between fluctuating process conditions, ignition variablity, engine geometry, fundamental soot formation mechanisms, etc. An isolation facility such as the shock tube, therefore, is crucial for studying P M emissions, especially in conjunction with the HPDI technology development. For this preliminary study, the soot contribution from natural gas is the focus investigation, where methane-based fuels are tested under engine-relevant conditions. Premixed experiments with methane and methane/ethane blends are performed to in-vestigate the effects of combustion temperature, pressure, and equivalence ratio on soot emissions. Similar fuels and combustion parameters are also tested under non-premixed conditions using a high-pressure gaseous injector, where the additional parameter of in-jected fuel mass is included. The resultant soot emissions from each combustion event are measured with the P M sampling methodology developed in Chapter 3, with any dis-crepancies noted in the experimental procedures. The results from several experiment series (with various conditions kept: constant) for each flame type are analyzed to search 67 4.2. Premixed Study 68 for possible correlations between soot emissions and the relevant combustion conditions. Wherever possible, appropriate corrections due to contamination and particle losses are applied for accuracy and error analysis purposes, as well as to allow more direct and meaningful comparisons of the results. Finally, applicable conclusions to these prelimi-nary results are drawn along with suggestions for further research. 4.2 Premixed Study Due to initial uncertainties associated with small soot outputs from non-premixed methane experiments, premixed experiments (with relatively larger amounts of fuel) are performed to detect significantly higher black carbon signals above the contamination and instrument noise levels. The first experiment series involved only methane, under various equivalence ratios. A n ethane additive of 10% (mole fraction) is then used to resemble the composition of natural gas. This allowed higher soot emissions for com-parison purposes, as well as to aid the particle loss studies mentioned earlier. In each series, either the combustion temperature or pressure is varied while keeping the other parameters as constant as possible, in order to investigate possible correlations between the total soot mass and the corresponding variable. The third series includes blank tests directly before and after the various premixed combustion runs, under similar experimen-tal conditions. This is primarily used to establish an appropriate baseline black carbon level to correctly interpret and analyze the measured soot amounts. Although there are only a limited number of experiments in each series, the results from these preliminary premixed tests can be used to compare against and complement the more substantial set of non-premixed experiments using similar fuels. 4.2.1 Apparatus The main components of the shock tube apparatus are previously described in Section 3.3. The optical section is not used due to high peak pressures in the premixed combustion environment. The double diaphragm system used to safely control the driver-to-driven pressure ratios consists of two identical diaphragms in series, separated by a small chamber (~ 1% of the driver section volume). This cylindrical void is charged to an intermediate pressure so that neither diaphragm wil l reach its burst pressure, until the chamber pressure is manually released (to atmosphere) by a solenoid valve. By 4.2. Premixed Study 69 choosing the appropriate diaphragm type and thickness so that only 80% of the burst pressure (Table B. l ) is reached, random diaphragm variability due to material and/or manufacturing defects can be avoided. The tube is fitted with five P C B Piezotronics 112B11. dynamic pressure transducers flush-mounted along the driven section (at 1.363 m, 3.195 m, 3.793 m, and 3.935 m with the closest one to the diaphragm section taken as the zero reference location) to detect the passage of the incident shock wave. A n Auto Tran 600D-117 vacuum sensor was used to prepare driven gas compositions and measuring initial driven gas pressures, while the driver gas pressures were measured by an Eclipse high-pressure sensor. Both pressure sensors were connected to a Circuit-Test DMR-3600 multimeter, where the corresponding voltages were measured and recorded. Outputs from the dynamic pressure transducers were recorded by a Wavebook/512 data acquisition system set to sampling frequencies between 125 kHz and 140 kHz (depending on the number of active channels used), corresponding to sampling intervals of 8 and 7.1 / is for each channel. Two photomultipliers (Electron Tubes P30A and TSI 9162) were used to monitor. C H (bandpass filter of 470±15 nm) and broadband light emissions, respectively, by placing them against the driven section endplate through a small quartz port. These optical light signals are also recorded by the data acquisition system, and used to detect contamination particle ignition as well as main fuel combustion. Steel shim diaphragms are used for all three experimental series, with the appropriate thickness (0.006"-0.008") determined from the required burst pressures needed to achieve the desired experimental conditions. Figures 3.1-3.3 show the shock tube apparatus and schematic. After each experiment, the particulate matter sampling system apparatus (shown in Figure 4.1) is used to measure the total black carbon emissions from the combustion event. As mentioned earlier, a non-surface-conductive polyethylene sample bag is initially used to collect the total shock tube contents, by rapidly venting through a ~" ball valve. A dual-wavelength Aethalometer (Magee Scientific AE21) is used to interpret black carbon masses from optical attenuation measurements. It continuously draws (using its internal pump and flow meter) the gas sample from the bag, traps the particles in the analysis chamber, and vents the rest of the inlet stream to the atmosphere. The developed multi-magnet filter is placed between the bag and the Aethalometer to trap possible steel diaphragm fragments and shock tube surface shavings. Surface conductive (silicone-coated) aerosol sampling tubing is used from the magnetic filter to 4.2. Premixed Study 70 the Aethalometer, while stainless steel tubing is used in the rest of the system in Figure 4.1. Aethalometer F igu re 4.1: Premixed Particulate Matter Sampling Apparatus Static pressure calibration was not performed on the dynamic pressure transducers along the driven section, as they are only used to detect shock wave passage. Previous dynamic calibration [21] shows dynamic response time of less than 3 / t s and a voltage-to-pressure conversion factor of approximately 930 ps i /V . The vacuum sensor used to measure initial driven section pressure was calibrated using a zero and span calibration corresponding to vacuum and atmospheric pressures prior to each experiment. The reference atmospheric pressure is taken from an Oakton Aneroid barometer. Linear regression between these calibration points was found to agree well with the manufacturer specifications [21]. The driver section high-pressure transducer was previously calibrated [19] to give a pressure conversion factor of 1250 ps i /V, which is close to the manufacturer data. Calibration on the Aethalometer beam signal voltages are performed automatically by the instrument upon startup, at the beginning of each day, and after every tape spot advance. The internal mass flow meter is manually calibrated as described in Appendix F. 4.2.2 Procedure The experimental procedures' consist of shock tube operation and P M sampling. Prior to each experiment, barometric pressure was recorded and the driven and driver gas pressure transducers calibrated. The driver and driven gas compositions required for each set of desired experimental conditions are determined by a numerical model [19], 4.2. Premixed Study 71 since the incident shock speed required to achieve the tailored interface depends on the target experimental conditions. These gas amounts are combined with the calibration pressures to determine the various voltages (of each gas species) required in each section of the tube. The entire tube is evacuated using a Cenco Instruments Hyvac 7 vacuum pump, before the calculated fuel(s) (Praxair Grade 2.0 ethane, Praxair Ultra High Purity methane) and then air (Praxair Medical Grade) amounts are added to the driven section, using linear interpolation of the calibration voltages. To enhance mixing in the methane series, the fuel and air are added in six alternating steps. This driven gas mixture is then allowed to settle for approximately 60 minutes to further promote mixing. The total driven pressure is measured again after the mixing period to check for possible leakage of room air and/or contaminants into the driven section. The data acquisition system is then armed (to be triggered by the rising edge of the incident shock wave as it passed the first dynamic pressure transducer). The driver gas composition is next prepared manometrically by first adding the required air amount, before filling the balance of the calculated driver pressure with helium (Praxair Industrial Grade). After approximately 50% of the final driver pressure is reached, the intermediate diaphragm chamber is isolated with a valve to allow equal pressure distributions (~ 80% of burst pressure) between the two diaphragms. When the final driver pressure is reached, this gap pressure is released to allow diaphragm rupture and shock wave initiation. Incident shock velocities were calculated using the pressure traces from four dynamic pressure transducer signals, by linearly fitting the time intervals (between rising edges) against their relative locations (as shown in Figures G . l and G.2). Using the measured incident shock velocity and initial driven gas properties, the pressure and temperature behind the reflected shock wave (i.e. experimental conditions) were calculated with isentropic normal shock relations and ideal gas (with variable specific heats) assumptions, using the same numerical model [19]. The light emission signals from the photomultipliers were analyzed for maximum and integral light intensities, to aid in the detection of particle contamina-tion as well as the level of combustion intensity as it relates (possibly) to soot production. As most of the developed particulate matter sampling procedures were described in detail in Chapter 3, they will be briefly summarized here with applicable deviations noted. Before each experiment, the driven section is cleaned with the supersonic multi-jet tool (using filtered Praxair air at 80 psi). The entire tube is kept under a positive pressure environment to prevent particle entry. The driver section is also cleaned periodically 4.2. Premixed Study 72 (~ every 3 experiments) with the supersonic air jets, as some diaphragm pieces are occasionally carried by the reflected shock wave back into the driver section. Whenever the tube surface gets noticeably sooty (e.g. after fuel-rich experiments), the steel brush and solvent cleaning procedures are used for the driven section, before it is rinsed with water and blown dry with a steady flow of filtered air. Blank tests are performed after each thorough cleaning procedure to confirm that background levels are similar to those before the sooty experiment. Prior to each experiment, any remaining bag contents are withdrawn with a small pump, before the bag is connected to the venting valve (similar to Figure C . l l ) . Immediately after the shock wave trigger, the entire shock tube contents were vented with the ball valve fully open into the sample bag (without particle impaction and tube settling). A non-conductive high-density polyethylene bag surface is used for all three series, with the electrostatic losses accounted by using only the first few minutes of Aethalometer data. The bag is then connected to the Aethalometer without a settling period, with the magnetic filter placed directly downstream of the | " bag fitting. The bag is periodically filled with filtered helium (under similar shock tube venting conditions) to check for possible particle accumulation. If significant bag contamination (above baseline tube bypass levels in Table 3.1) is detected, the bag is flushed several times, with any further necessary cleaning done by wiping the noticeable surface particle spots with a damp cloth. Similar procedures in the shock tube and P M sampling system operation are used in the blank tests for consistency, where air is used for the entire calculated driven section pressure. The Aethalometer is operated at one-minute timebase periods, with the U V channel turned off (due to the relatively rapid U V P M loading). The initial attenuation value immediately before sampling is recored as an indication of the tape spot 'freshness', which can be used to help compensate for the loading-induced scattering effect. The reference beam fluctuation level is also recorded for stability monitoring purposes. The total black carbon mass from each experiment is calculated using the modified data algorithm described in Section 3.7. The minute-to-minute attenuation values are first used to calculate corresponding increments of particle mass collection on the tape, using the specific attenuation factor (a) of 16.6 m 2 / g and a measured tape spot area of 0.5 cm 2 . These mass increments are then plotted against the time lapse from the sampling start time. The representative mass increment per minute is selected (with some operator discretion as shown in Section 3.7) as the first point after the Aethalometer has stablized 4.2. Premixed Study 73 from inlet gas composition changes. Using the initial driver and driven pressures in the shock tube, the initial total volume of gas in the sample bag is calculated, along with the true sample flow rate obtained from interpolations of helium percentage and initial mass flow meter setting. Therefore, the time needed to sample the entire bag can be found. Finally, the representative mass increment is combined with the total sampling time to arrive at the total B C mass from the experiment. The quartz tape is manually advanced to a new spot before the start of each experiment, along with its automatic beam and flow signal calibration and initialization sequence. A l l P M sampling data (and the instrument log file) from the Aethalometer is stored in the internal floppy disk, before being post-processed according to Section 3.7. 4.2.3 Results and Discussion The target experimental test matrices are shown in Tables G.1-G.3 in Appendix G , along with the summary of actual experimental conditions in Tables G.4-G.6 and Aethalometer measurements in Tables G.7-G.9. The relevant Aethalometer parameters are recorded to possibly help interpret the measured black carbon quantities. The relative timing of the blank experiments is included in Table G.6, in order to check for possible contamination against adjacent combustion experiments. The total and normalized black carbon masses are calculated and summarized in Tables G.7-G.9. Fuel masses were used for the normalization in combustion experiments while a common sample volume of 350 liters (typical of 30 bar pressures) was used to normalize blank experiment results. Due to the inherent apparatus limitations in achieving the predicted combustion conditions, certain experiments are neglected in this preliminary analysis and discussion. Slight variations in some dependent variables will also be neglected at this stage, as the limited data do not support three-dimensional (or higher) surface plot correlations. Similarly, correlations in variables with insubstantial data points will not be attempted, with the corresponding variable range combined into one series instead. For example, small pressure variations will be ignored since it is known to insignificantly affect soot production. Figure 4.2 shows the normalized B C mass results as a function of experimental tem-perature and equivalence ratio for the methane series. Due to the few data points for equivalence ratios above and below 1.0, their results are combined. It can be seen that 4.2. Premixed Study 74 120 100 3 O CO "a a) N 15 E L_ o 80 60 40 20 — — A . A : j — • A • EQR=1 A EQR<1 » EQR>1 A 9 " 1 1 • 1—• 1 1 1020 1070 1120 1170 1220 Temperature (K) 1270 1320 Note. Pressures for each series are between 22 and 31 bars F i g u r e 4.2: Premixed Methane Series Results the results show a significant scatter with no clear dependence on either variable. The extremely high B C masses from the fuel-lean (EQR < 1) experiments are unexpected. These are likely due to the entire driven section being soaked in Crystal Simple Green solution for a few days, prior to the relevant tests. Although it was thoroughly rinsed, the residue solvent on the surfaces could add to the actual B C from the fuel. The apparent decrease from experiment M3 to M9 seem to coincide with the carbon shim diaphragm trap (Figure B.5) and thus indicating significant diaphragm fragment contamination in the measured B C signals. The results from the fuel-rich (EQR > 1) experiments seem to be close to those under the stoichiometric conditions. Although this could be due to the 'cleanness' of the methane fuel, more data points are needed for a meaningful comparison. Figure 4.3 show the corresponding results for the methane/ethane series, with a 10% ethane mole fraction to resemble natural gas. Although the data is separated into five series, there are very few points in each to be used for analysis. Relatively small pressure variations are again ignored, as well as two extremely high B C results stated. It can be seen that the expected higher B C mass with increasing E Q R is not present, 4.2. Premixed Study 75 60 50 f m in TO 5 O m T3 0) N 15 E-40 30 20 10 1000 EQR=1 A EQR=0.5 • EQR=2 [xEQR=3 + EQR>3 X 9 1050 1100 Temperature (K) 1150 1200 Note: 1) Pressures for each series are between 34 and 42 bars 2) Points not shown on the graph for clarity: 207 ug/g (EQR=1), 303 ug/g (EQR=3) F igu re 4.3: Premixed Methane/Ethane Series Results as well as any temperature dependence. Comparing against the methane results, the addition of ethane does not clearly., increase soot production. The significant scatter of the results is again likely from the contamination of adjacent experiments, which is difficult to control in the relatively sooty premixed environment. For example, increasing the E Q R in experiments ME8 to ME12 seem to drastically increase the soot mass (with a visible soot layer coating on the tube walls) after a 'critical' E Q R is reached, resulting in the relatively high ME12 data. However, once this sooty condition is reached, it is very difficult to restore the tube to its 'clean' state, where the subsequent EQR=1 experiment also resulted in extremely high masses. Only after thorough cleaning of the tube, bag (wiping inner surfaces), and all sample flow paths did the B C mass return to pre-sooty levels. Since cleaning introduces combustible solvent to the environment, it will compound the contamination control problems. Figure 4.4 shows the B C mass results of the blank experiments as a function of temperature and pressure. The previous, combustion experiment results (without 4.2. Premixed Study 76 30 25 D) 3 20 O m -u <u N re E i_ o 15 10 V X • 15 +/-1 bar A 25 +/- 3 bar «41 +/- 3 bar x Methane + Methane/Ethane + * X x • -tx x x X • * + * X -I ± , , , + + x 0 1000 1050 1100 Temperature (K) 1150 1200 Note: 1) BC mass results are normalized to a common sample volume of 350 L 2) Blank test results are separated into three pressue ranges 3) Points not shown on the graph for clarity: 83 ug, 106 ug, 318 ug (methane/ethane) Figure 4.4: Blank Test Results for Premixed Series EQR differentiation) are also added for comparison and to help detect pre- and post-experiment contamination sources and levels. Al l BC masses are normalized to typical 30-bar experiment sample volumes of 350 liters. For the blank tests, temperature or pressure dependences are not evident. However, the two combustion series clearly produce more soot than the blank series, with the exception of a few methane/ethane experiments at stoichiometric conditions. In addition, the largest normalized BC masses (not shown for clarity) result from the methane/ethane fuel under high EQR, which is also expected. Nevertheless, due to the relatively small number and relatively large scatter of the blank series data points, it is difficult to contrast the quantitative results to the other two series. To investigate the possible contamination contribution in Figures 4.2 to 4.4, the relative timing of the high soot-producing blank experiments can be used (in Table G.6). It can be seen that the high blank BC mass experiments (Bl , B2, B3, B8, B9) are performed immediately after sooty runs, where the leftover soot particles on the tube walls could not be accessed by cleaning but instead is loosened by blank runs. 4.2. Premixed Study 77 This significant contamination is clearly detected in the relevant blank run B C results and must be controlled. In addition, the introduction of Lexan diaphragms (starting in B12) seem to consistently achieve lower measured B C values, in the absence of sooty experiments. Therefore it is advised to use Lexan for all subsequent experiments, thus allowing the remaining steel diaphragm fragments to gradually empty the tube over time. This removes the added complexity of magnetic filtering, as well as eliminating optical signal contamination from steel particles. Even if the Lexan particles are slightly charred in the flame, its optical absorption is far less than that of the opaque carbon or stainless steel material. It is important, however, to note the preliminary nature and the limited number of these premixed experiments, and to interpret their results accordingly. The lack of discernable trends of the B C masses with any combustion parameters is likely due to particle con-tamination. The relative timing of the experiments is important as a sooty run can take a few subsequent shock waves to restore the 'clean' tube surfaces, especially prior to the mechanical honing. Since shock waves are known to be extremely effective at entraining particles inaccessible by mechanical cleaning [45], it should be utilized to remove residue particles, before it is lost again to the tube surface due to various aerosol dynamics (espe-cially in high concentration environments as discussed in Section 3.6). It is also possible that particle deposits/patches on the surface can reach a.critical depth before significant amounts are stirred up and airborne, thus making the accounting procedure more com-plex. This can be best observed by experiments (under similar conditions) performed in sequence, where gradual increases in B C mass can provide further clues. In addition, the high particle-loading of the Aethalometer tape spot also enhances an optical scattering effect, thus complicating the B C mass variations even further. Therefore, it is impractical to study premixed experiments at this stage of the sampling system development, without improvements in contamination control and particle loss minimization. Innovative clean-ing procedures must be developed for sooty runs, where the elusive nature of the soot particles must be carefully studied. Nevertheless, these premixed experimental results have been useful to explore the challenges in both particle and optical contamination controls, and provide a good foundation for improvements in those areas. 4.2. Premixed Study 78 4.2.4 Error Analysis For the premixed investigation, the combustion conditions (behind the reflected shock wave), fuel mass, equivalence ratio, and B C mass were influenced by various error sources, as discussed in Appendix G. Since the cumulative variable ranges were used to analyze the B C quantities, the corresponding error bars will be neglected in Figures 4.2 to 4.4. Statistical error analysis will also be neglected due to the limited number of experiments. D r i v e n G a s C o m p o s i t i o n The driven gas composition (air and fuel) was calculated from the partial pressures of each gas species added into the driven section of the shock tube. Contributions to this error include vacuum pressure, barometric pressure, and ambient temperature/humidity. Errors in the vaccum pressure transducer are due to the multimeter resolution limitations (±0.001 V and ±0.01 V for voltages less than and greater than 4 V , respectively) and calibration error (±0.001 bar in the barometer used for transducer calibration) [21]. The cumulative effect is an error of ±0.002 bar for each individual gas partial pressure. Although errors in ambient temperature and relative humidity affect the calculation of the required partial pressures (based on gas properties), their effects are relatively small and thus will be ignored. As shown in Appendix G, the resultant average errors are 1.84% in the methane mass and 0.011 in the E Q R for the methane series. The resultant errors for the methane/ethane series are 1.11% in the total fuel mass and 0.023 in the E Q R . E x p e r i m e n t a l T e m p e r a t u r e / P r e s s u r e The experimental temperatures and pressures are calculated from the driven gas compo-sition and pressure (discussed previously) and the incident shock velocity. The incident shock velocity is determined from the dynamic pressure transducer signals as shown in Figure G.2. A linear fit of the transducer locations against the passage times was used, where the radius of convergence exceeded 0.9998 for all experiments. The error in the incident shock velocity is therefore mainly attributed to the measurement errors in transducer separation distances and data acquisition system sampling rate. As shown in Appendix G , the resultant error in incident shock velocities is 4.68-7.57 m/s for all three series. By combining with the driven gas pressure errors, the cumulative uncertainties in the experimental temperature and pressure ranges from 7-16 K and 0.3-0.9 bars, respec- >• 4.2. Premixed Study 79 tively. The exact error values for a given experiment depend on the magnitude of the incident shock velocity and driven gas pressure. B lack C a r b o n M a s s Total black carbon masses from each experiment are calculated from the most represen-tative mass increment and total sampling time required. Uncertainties in the collection tape spot area, optical attenuation factor, and B C mass curve fitting process contribute to the mass increment error. Less dominant measurement errors in the total sample volume and flow rate result in the total sampling time error. As shown in Appendix G, output voltage errors (due to multimeter resolution) in the driver section high-pressure sensor of ±0.001 V (0.85 psi) and estimated external flow rate calibration errors of 2% are used for the total sampling time calculation. On the other hand, errors in tape spot diameter (±0.04 mm) and specific optical attenuation factor (±1.0) affect the B C masses more directly. Additional errors in the uneven distribution of particles over the illuminated aerosol area (e.g. outer edges) will be neglected. In summary, the combined total B C mass errors range from 9-11%, while the normalized B C mass uncertainties lie between 9-15% for the methane series and 10-13% for the duel-fuel experiments. Finally, errors in the selection of the most appropriate trend (and hence the most representative mass increment) in the B C mass curve (see Section 3.7) can also be introduced. Since this requires a certain degree of operator discretion, its contribution to the total B C mass error will not be quantified. Table 4.1 compares the standard errors for each series with the corresponding measure-ment errors. While the fuel-based normalized B C masses are shown, the total B C masses from each experiment are used in the error calculations to match the units of the blank test normalization. The standard errors for the three premixed series shown significant variability, as compared to the measurement errors. This is due to the large standard de-viations observed in the relatively small number of data points. The standard deviations are also comparable to or exceed the measurement averages, due to a few extremely sooty runs. It can also be seen that the normalized B C mass error is the dominant measurement error source. 4.3. Non-Premixed Study 80 Table 4.1: Comparison of Standard and Measurement Errors' Average Normalized BC Mass Number of Expts Standard Deviation Standard Error • aUfN Average Measurement Errors Fuel Normalized Mass (%) BC Mass (%) Premixed Methane (40.1 ug/g) 10.7 ug 15 9.1 2.3 1.8 11 Premixed Methane/Ethane (51.8 ug/g) 58.2 ug 13 120 33 1.4 11 Blank (350 L) 6.32 ug 17 6.0 1.5 n/a 9.4 Non-Premixed Methane/DME Blank (350 L) (53.5 ug/g) 0.161 ug 0.515 ug 8 -21 0.10 0.30 0.036 0.065 n/a n/a 10 12 Non-Premixed Methane Blank (350 L) (169 ug/g) 1.35 ug 1.20 ug 27 18 1.6 0.75 0.31 0.18 n/a n/a 9.5 9.5 Non-Premixed Methane/Ethane Blank (350 L) (243 ug/g) 0.299 ug 0.333 ug 19 9 0.21 0.35 0.048 0.12 n/a n/a 9.5 9.5 Note: Average BC masses are normalized by fuel mass (in parentheses) and by a nominal shock tube volume of 350 L 4.3 Non-Premixed Study The premixed investigation was useful in producing sufficient B C amounts to bring forth awareness in critical experimental challenges such as contamination control, particle losses, and instrument signal/data analysis. However, the primary project motivation of measuring and correlating detectable amounts of soot from injected gaseous fuels (e.g. in the HPDI technology) requires non-premixed investigations of natural gas-like fuels. These non-premixed experiments also allows an approach in the above technical chal-lenges from an opposite perspective, where the goal is to preserve the soot produced from the combustion event to enable its measurement in the Aethalometer. The first experiment series used a methane and dimethyl ether (DME) fuel mixture under varying temperatures and pressures, using the same injection duration. A pure methane fuel was then used to investigate a temperature dependence, using the same pressure and injection 4.3. Non-Premixed Study 81 duration. Lastly, a methane/ethane mixture was injected under a constant pressure and duration. Since blank tests to establish baseline levels are relatively more important in the low-BC environments of non-premixed experiments, three separate blank test sets (embedded in the corresponding series) will be performed. It is also important to note that the various non-premixed experiments are undertaken concurrent to the sampling methodology development; hence their differences in the relevant experimental proce-dures will be noted. In addition, the timing of the blank tests will also be included to help with B C data interpretation. Through the efforts of the non-premixed investigation, further knowledge and advancements in the key experimental issues are made, and are complemented with the premixed study to develop the sampling system methodology described in Chapter 3. 4.3.1 A p p a r a t u s Since the shock tube and sampling system apparatus has been described in Section 3.3 and 4.1, only the deviations from the premixed study setup will be mentioned here. An optical access section (Figure A.5) is attached to the end of the driven section to enable optical measurements and combustion visualization. As shown in the figure, it contained three fused quartz windows and one instrumentation window (for dynamic pressure transducer placement). The optical section also lengthened the driven section by 0.54 m, resulting in a total shock tube length of 7.90 m. The two photomultipliers (to monitor C H and broadband light emissions) are now attached to the bottom window of the optical section and aimed vertically upward at the tube centerline. For the methane and methane/ethane series, a Vision Research Phantom v7.1 CMOS-based camera (equipped with a 50 mm F/1.2 Nikkor lens) was used to image the entire combustion event (frame rate of ~ 31000 frames/second, with an effective integration time of 1 /xs per frame). For the methane/DME series, various original diaphragm materials (plastic transparency film, grooved stainless steel and aluminum [21]) as well as stainless steel shimstock were used, based on the target pressure. Lexan diaphragms of 0.030" thickness were used for the other two fuel series. As shown in Figure 4.1, the methane/DME series used a non-surface conductive polyethylene bag, with the magnetic filter to capture airborne metallic diaphragm fragments. However, for the other two fuel series, the sampling apparatus was further developed and shown in Figure 4.5, where the impactor is placed immediately downstream of the | " ball valve to maximize flow rate/velocity and 4.3. Non-Premixed Study 82 impaction efficiency. Other deviations include the use of a surface conductive (carbon impregnated) polyolefin bag and removal of the magnetic filter (while using Lexan diaphragms). Additional specific apparatus deviations (e.g. various impactor versions) applicable to each experiment will be noted in the experiment summary tables. Aethalometer F i g u r e 4 . 5 : Non-Premixed Particulate Matter Sampling Apparatus Fuel injection was performed by a Westport Innovations J-43 gaseous fuel injector clamped to the shock tube endplate as shown in Figure A.4. This magneto-restrictive in-jector enables rapid opening and closing times, with a 1.1 mm diameter precision-drilled nozzle at the tip (the methane/ethane series used a modified J-43 with a 275 /xm nozzle). Westport's WCut control software was used for controlling the injection duration and the delay between trigger signal detection and injector opening. The trigger signal was generated by the incident shock wave as it passed the last dynamic pressure transducer location (Figure G . l ) . Unless otherwise stated, the delay time was set to 0.2 ms to ensure injection began 100-800 /xs after the incident shock reflection off the endwall (i.e. after the experimental conditions have been achieved). To minimize fuel leakage into the shock tube before injection, a manual shutoff valve and an Advanced Fuel Components solenoid valve were installed upstream of the injector (Figure A.4). The manual valve is needed to seal the injector from the pressurized fuel inlet (connected directly to the fuel tank) between experiments. The solenoid valve was used to charge the injector immediately before injection, in order to minimize the charged duration of the injector and hence fuel leakage. In addition, this injector-charging solenoid valve was controlled by the same signal used for venting the diaphragm intermediate chamber. This was 4.3. Non-Premixed Study 83 necessary because any shock wave-based trigger would not provide sufficient opening time for the solenoid valve. Although a time delay (0-400 ms) can be used between the two solenoid valve openings to further minimize fuel leakage (by compensating for the intermediate chamber venting time) [21], it is not implemented due to various unpre : dictabilities. Finally, signals from the dynamic pressure transducers, injector control, and photomultipliers were all recorded by the Wavebook/512 data acquisition system. Pres-surized standard Praxair fuel mixtures of methane/DME (95%/5%) and methane/ethane (90%/10%), in terms of mole fractions, are used as the fuel sources in the respective series. In addition to the pressure transducer and Aethalometer calibration procedures men-tioned in Section 4.3.1, the J-43 injector is initially calibrated using methane mass flow rates and shown in Appendix H. This characterization is performed by injecting methane (under various reservoir pressures) into a test chamber of known volume, while monitor-ing the chamber pressure [20]. The injected mass at each pulse width (injection duration) is calculated from the pressure increase. The mass flow rates exhibit good linear corre-lation over the range of injection durations tested. Mass fluxes are normalized against the injection duration and pressure, with an average difference from theoretical values of 28%. This loss is due to the deviation from the ideal adiabatic and isentropic gas expansion model, with assumed choked flow (Ma = 1) at the nozzle. 4.3.2 Procedure There are several procedural differences in both the shock tube operation and P M sampling, compared to the premixed case. Before tube evacuation, the injector is manually triggered to inject any remaining internal fuel. The stainless steel tubing connecting the injector, injector-charging solenoid, and manual shutoff valve is then relieved to atmospheric pressure. While the manual valve (Figure A.4) is closed, the upstream inlet fuel line is pressurized to the desired injection pressure by opening and closing the gas cylinder valve. The driven section is subsequently filled with only air to the calculated pressure. The data acquisition system is then armed (to be triggered by the rising edge of the incident shock wave as it passed the first dynamic pressure transducer), with the injection delay and duration set through the WCut software. The driver section gas is subsequently composed manometrically with air and then helium. When the final driver pressure is reached, the manual fuel inlet valve is opened, followed 4.3. Non-Premixed Study 84 by the venting of the intermediate chamber to allow diaphragm rupture and shock wave generation. Simultaneous to this gap pressure-venting signal, the injector is charged to the desired injection pressure and fired after the set delay time. Incident shock velocities are again calculated from the dynamic pressure transducer signals as shown in Figures G . l and G.2. The resultant experimental pressure and temperature (behind the reflected shock) are again determined using the isentropic normal shock relations. Depending on the extent of the methodology development, experiments in each series are subjected to varying P M sampling procedures. For the methane/DME series, minimal . contamination control and particle loss steps were undertaken (except the use of magnetic filtering), due to the early stage of the overall project. In the other two series, similar cleaning procedures (as the premixed study) was performed, including supersonic air jets, positive pressure environment, and driver section cleaning. However, due to the small amounts of soot produced, steel brush and solvent cleaning of the driven section was done only after unexpected and noticeable particle increases in the tube and/or optical signals. The supersonic air jet cleaning of the driver section was also relatively less frequent (~ every 10 experiments), as residue Lexan fragments on the driver side are not expected to reach the combustion region in time. Blank tests after each distinct cleaning procedure are included within the experimental sets. The gas sample venting, Aethalometer connection, and bag cleaning processes are similar to those described earlier. However, due to the consistently low levels of measured B C , bag flushing and cleaning were less frequently required. Furthermore, the injector was neither pressurized nor connected to the fuel line for all blank tests, as well as not activating the injector software. Finally, the Aethalometer operating procedures are similar to those previously discussed. 4.3.3 Resu l t s and D iscuss ion The target experimental test matrices are shown in Tables H.1-H.3 in Appendix H, along with the summary of actual experimental conditions in Tables H.4-H.7 and Aethalometer measurements in Tables H.8-H.11. Similar to the premixed series, the relevant Aethalometer parameters are recorded to aid in the B C data interpretation. The blank experiments relevant to each series are listed in chronological order, in order to check for possible contamination against adjacent combustion experiments. The total and normalized black carbon masses are calculated and summarized in Tables 1 4.3. Non-Premixed Study 85 H.8-H.11. Fuel masses calculated from the J-43 injector characterization were used for the normalization in combustion experiments. To compare against blank experiments, a common sample volume of 350 liters (typical of 30 bar pressures) was used to normalize the measured B C masses. As mentioned previously, slight variations in some dependent variables will be neglected at this stage due to limited data. Similarly, correlations in variables with insubstantial data points will not be attempted, with the corresponding variable range combined into one series instead. For example, small pressure variations will be ignored since it is known to insignificantly affect soot production. in m O CO TJ a> N 120 100 80 60 40 20 800 A • 8+/-1 bar A11+/-1 bar • 17+/-1 bar x 23+/-2 bar X X • X • m r— 1 — 1 — , 1000 1200 1400 Temperature (K) 1600 1800 F igu re 4.6: Non-Premixed Methane/DME Series Injection Results Figures 4.6 and 4.7 show the normalized B C mass results as a function of experimental temperature for the methane/DME series. Due to the limited number of injection experiments and large dependent variable ranges, trends in B C mass with temperature and/or pressures cannot be observed in Figure 4.6. Figure 4.7 shows the normalized total masses from the blank experiments seem to be randomly scattered among those from the injection tests (under corresponding pressures). For the injection runs, a common sample volume of 350 L is used to convert the fuel-based normalization to 4.3. Non-Premixed Study 86 CO in in ro O CO XI V N 1300 1100 900 700 500 300 100 o 8+/-1 bar A11+/-1 bar © 17+/-1 bar X 23+A2 bar x B 8+/-1 bar oB 11+/-1 bar + B 23+/-2 bar - B 4+/-1 bar AO X + O o X + V i — - « - + — o o o o • x + t + , : 800 1000 1200 1400 Temperature (K) 1600 1800 Note: Prefix B indicates Blank Series F i g u r e 4.7: Non-Premixed Methane/DME Series Blank Results a volume-based normalization, in order to compare against the blank runs. Although the injected fuel should produce additional B C signals above the background, several explanations are possible for the observed results. First of all, the relatively small and 'clean' fuel used in this series is not sufficient to produce noticeable B C masses above the baseline levels. The timing of these experiments during the early stages of sampling methodology development (including the non-conductive bag) can also cause the variabilities in the results, where frequent deviations in apparatus and/or procedure exist (Table H.4). Since any foreign contamination introduced into the system requires subsequent shock waves and gas venting for its removal, a prolonged set of blank tests is likely required to restore the true background B C levels. In particular, variations in cleaning procedures for each experiment can greatly affect the amount of B C mass resulting from contamination particle combustion. This is evidenced, for example, in blank tests NMD17B and NMD18B, where inert driven gases (N2) are used to prevent combustion. As a result, their normalized B C masses lie in the lower end of the total range. Finally, due to the non-conductive bag surface used for this series, all mass increment curves exhibit a steady decay (similar to Figures E.2 and E.6), where the best 4.3. Non-Premixed Study 87 350 300 f 250 to in n3 5 200 O 3 150 N § 100 50 0 1100 ~ ~ ~~ 5 '• ; : . O 0 o o 0 ; O \* ' '. 1 o 1 , 1 ' 1 1 1150 1200 1250 1300 Temperature (K) 1350 1400 1450 Note: 1) All experimental pressures between 28 and 31 bar 2) Points not shown on the graph for clarity: 473 ug/g and 1026 ug/g Figure 4.8: Non-Premixed Methane Series Injection Results linear fit is used to obtain the most representative B C mass increment. The steady B C decrement can be attributed to the various particle loss mechanisms discussed in Section 3.6. In summary, the results from the methane/DME series clearly point to the need for cleaning improvements, sufficient B C generation, increased data at smaller variable ranges, as well as the importance of consistent experimental procedures and apparatus. Figures 4.8 and 4.9 show the normalized B C mass results as a function of experimental temperature for the methane series, where 30-bar experimental pressures are targeted. From Figure 4.8, there does not seem to be a clear temperature dependence on B C mass, with large variations occurring within very small temperature ranges. Similar to the methane/DME series, the normalized blank experiment masses (Figure 4.9) also show a fairly random scatter among the injection results. Instead of the expected lower results, blank tests over certain temperatures actually produced higher average B C masses. Nevertheless, there are several important observations from this series that can be used to further improve the sampling methodology. With the use of the surface conductive 4.3. Non-Premixed Study 88 2600 _ 2100 D) C tn co 1600 o CO T3 J 1100 CO -E 600 100 1100 — -O _ o Injection A Blank X o 0 o A A* I © * A 0 © _! *A A i 1 •—i 0 1150 1200 1250 1300 Temperature (K) 1350 1400 1450 Note: 1) All experimental pressures between 28 and 31 bar 2) Points not shown on the graph for clarity: 3962 ng (Injection), 8546 ng (Injection), 3563 ng (Blank) F igure 4.9:.Non-Premixed Methane Series Blank Results polyolefin bag (in addition to the advanced contamination control and cleaning proce-dures developed to-date), B C mass increment curve decay has been eliminated from blank tests. The blank test Aethalometer data exhibit relatively constant B C signals over the entire sampling period, with the remaining levels likely due to non-BC particles (e.g. diaphragm fragments) that do not experience high rate-loss mechanisms such as electrostatics. The observed linear decay in injection experiments can therefore be attributed to fuel-related B C with more confidence. Furthermore, the effects of inertial impactors are clearly evident in this series. From Table H.9, it can be seen that the use of the | " single-stage inertial impactor (starting at N M 2 2 B ) significantly reduced the normalized B C output, for both injection and blank experiments. In addition, the post-impactor B C mass increment curves exhibit more predictable constant (or slightly increasing) profiles, where the linear decay has also been eliminated for injection experi-ments. This is likely a result of the impaction removal of the larger particles that would otherwise have gradually settled in the bag. Finally, the trial usage of the three-stage 4.3. Non-Premixed Study 89 400 _ 350 D) 300 co O 250 •o CD N H 200 o 150 100 ~~ — © — © o © -© © © 0 © © © © © © 1100 1150 1200 1250 1300 1350 1400 1450 . 1500 Temperature (K) Note: 1) All experimental pressures between 29.1 and 30.4 bar 2) Points not shown on the graph for clarity: 875 ug/g Figure 4.10: Non-Premixed Methane/Ethane Series Injection Results impactor (starting at NM36B) proved ineffective, where both the B C masses and linear curve decay returned to pre-impaction levels (also confirmed by NM45B). This is almost certainly due to the insufficient flow velocities (for hypersonic nozzle conditions) in the latter impaction stages. Therefore the three-stage impactor design was subsequently discarded, with modified versions utilizing multiple nozzles to take advantage of a single high-pressure drop stage. Although the sampling methodology in the methane series (as compared to the previous series) was much more improved and consistent, challenges in blank test signal contamination, bag loss mechanisms/rates, and impaction dynamics still exist and wil l be explored in the next series. In particular, impactor design can be further improved (along with better knowledge of the relevant contamination species) to achieve the desired cutoff size and efficiency, without interfering with the path of B C particles. Figures 4.10 and 4.11 show the normalized B C mass results as a function of experimental temperature for the methane/ethane series, where 30-bar experimental pressures are 4.3. Non-Premixed Study 90 500 450 1? 400 O oo -o o> .b! "fo E L _ o 350 300 250 200 150 100 © o Injection A Blank © © © o © A © © A A A © A A © © © A % © © © 1100 1150 1200 1250 1300 1350 Temperature (K) 1400 1450 1500 Note: 1) All experimental pressures between 29.1 and 30.4 bar 2) Points not shown on the graph for clarity: 1084 ng (Injection), 1257 ng (Blank) F igu re 4.11: Non-Premixed Methane/Ethane Series Blank Results once again targeted. Ethane was introduced in an attempt to generate higher B C signals relative to the baseline levels. However, the amount of fuel injected was substantially less as the modified J-43 injector's nozzle diameter was reduced by a factor of 4. As shown in Table H.7, a single-stage multi-nozzle hypersonic impactor was used throughout this series to remove particles outside of the B C size range. A post-experimental tube settling, period is also added to filter out the observed blank test, B C signals in the previous series. At this stage, the length of the settling period is roughly estimated from previous settling trials as the time required to achieve blank test levels as shown in Table 3.1. As shown in Figure 4.10, there is no noticeable temperature dependence on the measured B C mass, with similar variations over the entire dependent variable range as the methane series. The expected increase in soot production due to the ethane additive is also not observed. However, the normalized blank test results in Figure 4.11 shows relatively smaller total quantities, as compared to the injection tests. This separation between the injection and blank results is likely due to the improvements in impaction and settling procedures, where the non-BC component of blank test particles are greatly reduced. For example, 4 .3 . Non-Premixed Study 91 cn in in co O CD 73 0) N "CO E t~ o 1700 1500 1300 1100 900 700 500 300 100 A_A o Methane/DME A Methane x Methane/Ethane A J X A o o 0 -o & - . • • — /. o o A AX —— y\—rt—— _ _ _ _ _ _ 9 : - « — o ** • . * 1 1 o o 1100 1200 1300 1400 Temperature (K) 1500 1600 1700 Note: 1) Methane/DME series experimental pressures between 4 and 24 bar 2) Methane series experimental pressures between 28 and 31 bar 3) Methane/ethane series experimental pressures between 29.1 and 30.4 bar 4) Points not shown on the graph for clarity: 2043 ng (methane), 3563 ng (methane) F i g u r e 4.12: Comparison of Non-Premixed Blank Experiment Results Table H . l l illustrates the effect of post-experimental tube settling, where an immediate decrease in blank test B C signals is observed. In addition, the use of a larger diameter plate with multiple hypersonic nozzles (starting in NME3B) allows the low blank levels to be consistently achieved, due to their combined individual impaction efficiencies. This obvious improvement over even the previous smaller diameter multi-nozzle impactor (e.g. N M E l B ) gives motivation to further impactor development, where the relative location and alignment of the nozzles as well as the nozzle and collector plates can be further investigated. In particular, variabilities in impaction dynamics due to the small size and proximity of the hypersonic nozzles can affect the desired cutoff size and efficiency, as well as possibly trap the soot particles. In terms of the various tube settling durations tested, there is no clear correlation to the measured B C signal over this limited data set. Therefore, further studies on the optimal tradeoff between contamination particle settling and soot particle loss to the tube walls should also be performed. The 4.3. Non-Premixed Study 92 use of settling, conductive bag, and improved impaction in this series has also produced more predictable B C mass increment, curve profiles. The relatively constant (or slightly increasing) mass increments from blank tests have been maintained from the previous series. The injection experiments also produced fairly constant profiles in most cases, with a very small linear decay in a few tests. This slight decrease could be due to the various bag loss mechanisms, in cases where the applicable particle species escaped the shock tube. The remaining variabilities in the mass increment curves, however, give an indication of the areas for further sampling methodology development. The key is-sues to be resolved include contamination prevention and soot particle loss minimization. Finally, overall comparisons of blank test results from all three series (irrespective of pressures) are shown in Figure 4.12 as a function of temperature. Corresponding com-parisons using the injection test results are not attempted due to the variations in fuel species and mass flows. It can be seen that the methane series (after thorough tube surface treatment/cleaning, pre-experimental contamination control, Lexan diaphragm and conductive bag usage) did not result in consistently lower blank B C levels. However, the introduction of particle impaction and settling in the methane/ethane series was clearly effective in reducing the detected B C signals from blank runs. Therefore these technical areas should be investigated further to achieve a minimal baseline level, before corresponding injection experimental results can be given more meaning. Furthermore, higher B C masses should be produced (using larger injection flow rates and sooty fuels) to aid in the proper interpretation of the fuel specific B C output. These steps will lead to the eventual attribution of soot production as a function of the combustion variables investigated. 4.3.4 Error Analysis For the non-premixed investigation, the combustion conditions, injected fuel mass, fuel composition, and B C mass were influenced by various error sources. The combustion con-ditions and Aethalometer B C mass results are affected by similar errors as the premixed study and previously discussed in Section 4.3.4. Error bars corresponding to combus-tion conditions will be neglected in Figures 4.6 to 4.12, due to their cumulative ranges used to analyze the B C quantities. Statistical error analysis wil l also be ignored at this preliminary stage. Errors in fuel compositions wil l be neglected as standard Praxair gas 4.4. Conclusions 93 mixtures are used. Errors in the injected fuel mass are composed of injector leakage and mass flow calibration sources. Injector leakage can occur before the actual injection, while the manual valve in Figure A.4 is open. Although previous studies [21] estimate charged durations of 200-800 ms, the actual leaked fuel mass is difficult to determine. On the other hand, errors in injector mass flow rate calibration will also not be quantified due to its exclusion in the mass flow calibration raw data [20]. In addition, mass flow rates for the methane/DME and methane/ethane fuel mixtures are not corrected from the calibration data (using methane), due to the relatively small additive percentages and relevant property variations. As summarized in Appendix H, the cumulative uncertainties in the experimental temperature and pressure ranges from 11-15 K and 0.57-0.64 bar, respectively, depending on the magnitude of the incident shock velocity and driven gas pressure. As seen in Table 4.1, the cumulative B C mass error is still the dominant error source, and range from 9-18%, 9-10%, and 9-11% for the methane/DME, methane, and methane/ethane series, respectively. However, the standard deviations have been greatly reduced from the premixed series due to the increased measurement repeatability. This is also reflected (when combined with the larger data sets) in the smaller standard errors, and shows the progress of the sampling methodology development. 4.4 Conclusions The limited premixed investigation did not result in noticeable dependences of black carbon mass on combustion temperature, pressure, or equivalence ratio. Comparisons with normalized blank test results point to cleaning and sampling process inadequacies where particle and B C signal contamination are clearly present. The occasional sooty runs also obscured the results from adjacent experiments, and prevented the proper establishment of baseline levels in blank runs. Therefore it is critical to prevent high E Q R (and soot generation) tests, as well as rigorous manual and shock entrainment cleaning after the detection of visible,contamination. The non-premixed investigation incorporated contamination and particle loss reductions using conductive bag surfaces, lighter/transparent diaphragms, hypersonic impaction, and post-experimental tube settling. These improvements resulted in consistently achieving blank test B C levels of 100-300 ng, as well as more sensible injection and 4.4. Conclusions 94 blank experiment B C mass increment curve profiles. However, there were still no observed correlations between measured B C mass and combustion temperature or pressure (although a constant pressure was targeted in most cases). The use of tube settling was effective in clearly separating the injection and blank test B C output, thus possibly allowing the fuel-based B C to be properly accounted for and interpreted against background signals. Greater fuel masses can be injected to produce larger distinct B C signals, and provide meaningful fuel-specific sooting characteristics. The evolution of contamination elimination is also evident in the three series, and .illustrates the critical need for the various developed methodology. The dominant error source is again due to the specific attenuation uncertainty in B C mass measurement. Although the apparatus and procedural variabilities during the sampling methodol-ogy development partly contributed to the observed scatter of the results, inherent variablities in ignition [47] also exist. The post-experimental contamination control ' and electrostatic loss minimization are found to be the most effective. Compared to pre-experimental cleaning and entry prevention, particle impaction and settling are much more effective in eliminating contamination B C signal detection. Soot particle electrostatic losses are also greatly minimized by a highly surface conductive bag. In terms of Aethalometer data analysis, the entire time-dependent mass increment profile should be used to understand each experiment-specific process and to calculate the most representative value. Even though settling durations of 30-40 minutes can generally achieve the temporally flat B C mass readings (i.e. remove the observed particle loss/decay), some discretion in the curve fitting should be used at this stage without entirely understanding the particle loss physics. For example, data near the end of the sampling period (with small distances to bag surfaces) should be discarded due to the expected (and observed) higher losses. In addition, the U V channel can theoretically be used to determine quantities of non-BC material present, thus further aiding the 'true' B C mass accounting process. The 2007 U.S. diesel-engine P M regulations (Figure 2.1) of 0.01 g/bhp-hr equates to approximately 70 ng per mg of natural gas (or 70 ug/g). Although some experiments (in both studies) meet this standard and suggest the suitability of methane as a clean fuel, the relative standard deviations (Table 4.1 are too high at this stage to justify any quan-titative comparisons. It is recommended to further improve the sampling methodology 4.4. Conclusions 95 in areas of contamination, particle loss studies, impactor design, and B C mass interpre-tation. The use of a rapid acting sampling valve (with a large opening) to coordinate with the experimental timing should be studied, given the known effectiveness of ther-mopheoretic tube surface losses as well as the shock wave entrainment of contamination particles. The relevant particle loss mechanisms throughout the sampling path (including the impactor plate) should be quantitatively analyzed, especially the soot/contamination loss tradeoff during settling. The required impaction cutoff size and efficiency should be investigated through the use of the C P C / D M A (for size and quantity distribution) and S E M / T E M analysis (for particle properties such as shape and density). The particle types and sizes of typical gas samples should also be used to determine the applicable shock tube-specific optical specific coefficient. In addition, light emissions, video footage, and laser-based diagnostics can potentially be useful external checks on the measured B C amounts. Finally, larger sets of experiments should be performed to validate the sampling methodology. These should include larger engine-relevant variable ranges (e.g. distinct high and low pressures), under both premixed and injection environments. Chapter 5 Conclusions and Recommendations The measurement of particulate matter from shock tube combustion has been motivated by the stringent emissions standards, where natural gas-based fuels are implemented in diesel-cycle engines to realize the potential benefits. To investigate the possible correlations between soot emissions and combustion variables, the shock tube is used due to its attractive features outlined earlier. A series of premixed and non-premixed experiments using methane-based gaseous fuels was performed. Although the goal of accurate and repeatable B C measurements under engine-relevant conditions remains to be validated, the results of this study provide a valuable foundation for future work. The sampling system development focused on experimental conditions attainment, contamination control, particle loss minimization, and proper instrument detection. Contamination is detected by light emissions and verified through microscopy techniques. Rigorous pre- and post-experimental contamination controls are needed to ensure a background B C level can be established in blank tests. Compared to tube cleaning and particle entry prevention, hypersonic particle impaction and tube settling are found to be more effective. Potential soot particle losses (through each sampling path) via thermopheoresis, gravitational settling, diffusion, and electrostatics are qualitatively considered. Specifically, conductive tube and bag surfaces are highly effective in minimizing electrostatic losses, where it is especially important for the smaller sized soot particles. Further studies of the physics of loss mechanisms, including quantitative percentage accounts, should be performed. Aethalometer data algorithms are modified to suit the needs and limitations of the current setup. Total B C masses are determined using the most representative mass increment (from the manual data processing) and 96 5. Conclusions and Recommendations 97 the empirically measured sample flow rate. Due to the dominant uncertainty in the specific attenuation coefficient, external validation methods should be undertaken. The combination of these methods has resulted in blank test levels of 100-150 ng. Although the remaining background B C signal is due a combination of sources, it can be produced by such a minute amount of aerosolized particle that it is impractical to eliminate. The preliminary results from premixed and non-premixed methane-based flames demon-strate the sampling methodology, and provide a crucial tool in illustrating the unresolved challenges. The limited set of premixed experiments did not result in noticeable B C trends with combustion temperature, pressure, or equivalence ratio. High E Q R tests should be prevented due to the persistent cleaning and shock wave entrainment required to re-establish the appropriate baseline. Even though the more substantial set of non-premixed experiments also showed no clear dependences on temperature or pressure, much more sensible B C mass increment curve profiles are consistently obtained using a combination of conductive bag, impaction, and settling. The total blank B C levels are also kept between 100-300 ng, which allows greater injected fuel masses to generate clearly distinguishable signals. It was found that post-experimental contamination control and electrostatic loss minimization are the most critical procedures. Particle impaction and settling are also more effective (compared to pre-experimental cleaning and entry prevention) in eliminating contamination B C signal detection. The revised Aethalometer data interpretation is also necessary to suit the current study objectives. Errors can be attributed to uncertainties in measurements, particle dynamics, and B C signal interpretation. Although each individual error can conceivably be accounted for in the final results, their unknown interactions make it extremely difficult to quantify at this stage. However, the preliminary results do show the promise of methane (and potentially nat-ural gas) as a clean fuel, as compared to the 2007 diesel engine P M standard equivalent of 70 ng/mg of natural gas. It is recommended to validate the sampled B C mass using an independent method, as a priority. Long-term efforts using liquid fuels, duel fuels, and E G R simulation can be useful to compare against corresponding engine results, and determine the practical limits of this sampling system. On the other hand, N O x and gas chromatography studies using this sampling system can provide complementary under-standing of shock tube combustion. Finally, it should be kept in mind that combustion 5, Conclusions and Recommendations 98 and soot physics are in itself complex phenomena. The natural variability of ignition (location, delay), flame propagation (detonation, deflagration), chemical kinetics (rates, species), and particle dynamics (process interactions) makes a single combustion event difficult to compare to representative average values from engine tests. Although much more work remains to validate this sampling system and methodology, it is extremely difficult to physically measure the minute amount of P M emitted from methane flames in a shock tube. Therefore it is more practical to further investigate the emissions reduction potential of natural gas under engine environments, and thus contribute to the overall alternative fuels efforts. References [1] A . Alexiou and A. Williams. Soot formation in shock-tube pyrolysis of toluene-n-heptane and toluene-iso-octane mixtures. Fuel, 74(2):153-158, 1995. [2] A . Alexiou and A . Williams. Soot formation in shock-tube pyrolysis of toluene, toluene-methanol, toluene-ethanol, and toluene-oxygen mixtures. Combustion and Flame, 104:51-65, 1996. [3] G.A. Allen, J . Lawrence, and P. Koutrakis. Field validation of a semi-continuous method for aerosol black carbon (aethalometer) and temporal patterns of summer-time hourly black carbon measurements in southwestern pa. Atmospheric Environ-ment, 33(5):817-823, 1999. [4] D. Baumgardner, G. Raga, O. Peralta, I. Rosas, and T. Castro. Diagosing black carbon trends in large urban areas using carbon monoxide measurements. Journal of Geophysical Research, 107(AAC):X1-X9, 2002. [5] K. Brezinsky. High Pressure Shock Tube Gas Sampling. Uni-versity of Illinois at Chicago. Personal Communication, 2003, / ~kenbrez/html / sampling.html. [6] P. Cadman. Shock tube combustion of liquid sprays of ethanol: Effect of additives on ignition delays and combustion characteristics. 20th International Symposium on Shock Waves, 1:1039-1045, 1996. [7] P. Cadman and R.J . Denning. Oxidation rates of soot particulates by oxygen in the temperature range 1500-3500K determined using a shock tube. Journal of the Chemical Society, Faraday Transactions, 92:4159-4165, 1996. [8] P. Cadman, R . J . Denning, and I.L. Morris. Oxidation rates of soot particulates in the presence of hydrogen/oxygen mixtures at high temperatures. 21st International Symposium on Shock Waves, 1(1690), 1997. [9] J . Chomiak. Current approaches toward a clean diesel engine. 20th International Congress on Combustion Engines, (D10), 1993. 99 R E F E R E N C E S [10] H.J. Cohen, G. Sirianni, S. Chemerynski, J . Borak, and R. Wheeler. Observations on the suitability of the aethalometer for vehicular and workplace monitoring. Journal ,of Air & Waste Management Association, 52:1258-1262, 2002. [11] D.F. Davidson, M.A. Oelschlaeger, J.T. Herbon, and R.K. Hanson. Measurements of iso-oetane ignition times and oh concentration time histories. 29th Symposium (International) on Combustion, 1:1295-1301, 2002. [12] American Society for Metals. Metal Handbook. Desk edition, Metals Park, O H , 1985. [13] I. Glassman. Combustion. Academic Press, second edition, Orlando, 1987. [14] A . D . A Hansen. The Aethalometer. Magee Scientific Company, Berkeley, 2003. [15] J .B . Heywood. Internal Combustion Engine Fundamentals. McGraw-Hi l l , New York, 1988. [16] W . C . Hinds. Aerosol Technology: Properties, Behaviour, and Measurement of Air-borne Particles. John Wiley k. Sons, second edition, New York, 1999. [17] H. Horvath. Experimental calibration for aerosol light absorption measurements using the integrating plate method - summary of the data. Journal of Aerosol Science, 28(7): 1149-1161, 1997. [18] C.H.. Huang, C . J . Tsai, and T.S. Shih. Particle collection efficiency of an inertial impactor with porous metal substrates. Journal of Aerosol Science, 32:1035-1044, 2001. [19] J . Huang. Experimental shock tube study of ignition promotion for methane under engine-relevant conditions. Master's thesis, University of Brit ish Columbia, 2001. [20] J . Huang. J-43 Injector Mass Flow Characterization. Personal Communication, 2002. [21] J.L. Iaconis. A n investigation of methane autoignition behaviour under diesel engine-relevant conditions. Master's thesis, University of Brit ish Columbia, 2003. [22] Omega Engineering Inc. The Flow and Level Handbook. Stamford, C T , 1990. [23] Sierra Instruments Inc. Sierra 820 Series Top-Irak Mass Flow Meter Instruction Manual. Monterey, C A , 1994. [24] T.P. Jenkins and R.K. Hanson. Soot pyrometry using modulated absorp-tion/emission. Combustion and Flame, 126:1669-1679, 2001. [25] H. Jones. Particulate matter emission sources in hpdi engines and strategies to reduce pm emissions. Master's thesis, University of Brit ish Columbia, 2004. R E F E R E N C E S 101 [26] H. Jung, D.B. Kittelson, and M.R. Zachariah. Kinetics and visualization of soot oxidation using transmission electron microscopy. Combustion and Flame, 136:445-456, 2004. [27] H. Kellerer, R. Koch, and S. Witt ig. Measurements of the growth and coagulation of soot particles in a high-pressure shock tube. Combustion and Flame, 120:188-199, 2000. [28] H. Kellerer, A . Muller, H.J. Bauer, and S. Witt ig. Soot formation in a shock tube under elevated pressure conditions. Combustion Science & Technology, 113/114:67-80, 1996. [29] I.M. Kennedy. Models of soot formation and oxidation. Progress in Energy and Combustion Science, 23:95-132, 1997. [30] R.D. Kern and K. Xie. Shock tube studies of gas phase reactions preceding the soot formation process. Progress in Energy and Combustion Science, 17:191-210, 1991. [31] E.O. Knutson and K.T. Whitby. Aerosol classification by electric mobility: Appara-tus, theory, and applications. Journal of Aerosol Science, 6:443-451, 1975. [32] C. Kopp, A. Petzold, and R. Niessner. Investigation of the specific attenuation cross-section of aerosols deposited on fiber filters with a polar photometer to determine black carbon. Journal of Aerosol Science, 30(9): 1153—1163, 1999. [33] A . Kunz, R. Wang, and P. Cadman. Liquid spray combustion of propanol/tetradecane/water mixtures. 21st International Symposium on Shock Waves, 1(1691), 1997. [34] L .E. LaRosa, T . J . Buckley, and L.A. Wallace. Real-time indoor and outdoor mea-surements of black carbon in an occupied house: An examination of sources. Journal of Air & Waste Management Association, 52:41-49, 2002. [35] C. Liousse, H: Cachier, and S.G. Jennings. Optical and thermal measurements of black carbon aerosol content in different environments: Variation of the specific attenuation cross-section, sigma (CT). Atmospheric Environment, 27A(3):1203~1211, 1993. [36] I.G. Loscertales. Mass diameter versus aerodynamic diameter of nanoparticles; impli-cations on the calibration curve of an inertial impactor. Journal of Aerosol Science, 31:923-932, 2000. [37] A . Muller and S. Witt ig. Influence of temperature and pressure on soot formation in a shock tube under high pressure conditions. Proceedings of the 18th International Symposium on Shock Waves, 2:759-764, 1992: R E F E R E N C E S 102 [38] B.R. Munson, D.F. Young, and T .H . Okiishi. Fundamentals of Fluid Mechanics. John Wiley & Sons, third edition, New York, 1998. [39] C. Park and J.P. Appleton. Shock tube measurements of soot oxidation rates. Com-bustion and Flame, 20:369-379, 1973. [40] A. Petzold, C. Kopp, and R. Niessner. The dependence of the specific attenuation cross-section on black carbon mass fraction and particle size. Atmospheric Environ-ment, 31(5):661-672, 1997. [41] S.N. Rogak. Shock Tube Cleaning Notes. University of Brit ish Columbia. Personal Communication, 2003. [42] R. Said, A . Garo, and R. Borghi. Soot formation modeling for turbulent flames. Combustion and Flame, 108:71-86, 1997. [43] R.J . Santoro and J .H. Miller. Soot particle formation in laminar diffusion flames. Langmuir, 3:244-254, 1987. [44] S. Sidhu, J . Graham, and R. Striebich. Semi-volatile and particulate emissions from the combustion of alternative diesel fuels. Chemosphere, 42:681-690, 2001. [45] G.T. Smedley, D.J. Phares, and R.C. Flagan. Entrainment of fine particles from surfaces by gas jets impinging at normal incidence. Experiments in Fluids, 26:324-334, 1999. [46] R.E. Sonntag, C. Borgnakke, and G . J . Van Wylen. Fundamentals of Thermodynam-ics. John Wiley & Sons, fifth edition, New York, 1998. [47] G.D. Sullivan, J . Huang, T .X . Wang, W.K . Bushe, and S.N. Rogak. Emissions variability in gaseous fuel direct injection compression ignition combustion. SAE World Congress, 1(2005-01-0917), 2005. [48] C . J . Tsai, J.S. L in , S.G. Aggarwal, and D.R. Chen. Thermophoretic deposition of particles in laminar and turbulent tube flows. Aerosol Science & Technology, 38:131-139, 2004. [49] R. Wang and P. Cadman. Ceo detection in soot formed in benzene liquid spray combustion in a shock tube. Fullerene Science & Technology, 3:553-563, 1995. [50] R. Wang and P. Cadman. Soot and pah production from spray combustion of differ-ent hydrocarbons behind reflected shock waves. Combustion and Flame, 112:359-370, 1998. [51] J . Warnatz, U. Maas, and R.W. Dibble. Combustion: Physical and Chemical Fun-damentals, Modeling and Simulation, Experiments, Pollutant Formation. Springer-Verlag, third edition, New York, 2001. R E F E R E N C E S 103 [52] E. Weingartner, H. Saathoff, M . Schaiter, N. Streit, B. Bitnar, and U. Baltensperger. Absorption of light by soot particles: Determination of the absorption coefficient by means of aethalometers. Journal of Aerosol Science, 34:1445-1463, 2003. [53] P.O. Witze. Diagnostics for the measurement of particulate matter emissions from reciprocating engines. Fifth International Symposium on Diagnostics and Modelling of Combustion in Internal Combustion Engines, 1:7-12, 2001. [54] T. Zarutskaya and M. Shapiro. Capture of nanoparticles by magnetic filters. Journal of Aerosol Science, 31(8):907-921, 2000. Appendix A Background A . l Soot Formation Mechanisms A. 1.1 Soot Precursors In the absence of flame extinction, the hydrocarbon fuel breaks down to smaller hydrocar-bons such as C i , C 2 , and in particular C 2 H 2 (acetylene). The initial step in the production of soot is the formation of the first aromatic species from these smaller hydrocarbon frag-ments. The aromatic species grow by the addition of other aromatic and smaller alkyl species to form an important class of higher hydrocarbons called the polycyclic aromatic hydrocarbons (PAH) [51]. P A H compounds are usually formed under fuel-rich conditions, which is always present in non-premixed and rich premixed flames. The P A H formation process starts by C3H4 decomposition or reaction of C H or C H 2 with C 2 H 2 to form C3H3. After recombination to an aliphatic and rearrangement, the first ring (benzene: ) is formed as shown in Figure A . l , partly aided by the very slow competing oxidation reac-tions of C3H3. A n example of the elementary reaction mechanism for P A H growth from acetylene starts with the addition of C 2 H 2 to phenyl radicals, forming styryl radicals. A second C 2 H 2 adds to the styryl radical, and ring closure follows to form naphthalene. Further addition of acetylene to the ring leads to the growth of the P A H molecule. P A H growth can also be caused by aromatic structures as well. Although the details of this soot precursor formation is still debated, it is widely accepted that further growth of the P A H leads to the smallest identifiable soot particles with diameters and masses on the order of 1 nm and 1000 amu, respectively. Since soot precursors are pyrolyzed and oxi-dized at elevated temperatures, soot formation is limited to temperatures between 1000 104 A . l . Soot Formation Mechanisms 105 K and 2000 K. Figure A . l illustrates the precursor formation process. F i gu re A . l : Precursor Formation Process [51] A. 1.2 Particle Inception The production of soot particles in a flame is inherently a chemically controlled phe-nomenon. Thermodynamics alone cannot describe this process since soot is formed be-yond regimes where it is thermodynamically stable relative to the oxides of carbon. For example, hydrocarbon combustion in fuel-rich mixtures (with C O and H 2 as the main products) can be represented by: 777 CnHm + k02 - 2kCO + -H2 + ( n - 2k)Cs (A. l ) If soot formation is thermodynamically controlled, solid carbon should appear for n > 2k or a C / O ratio > 1 [51]. The corresponding fuel/air equivalence ratio is given by (A.2) where <p = 3 for C / O = 1, with m/n = 2 However, the experimentally observed critical C / O ratios are less than one, varying from 0.5 to 0.8 depending on the fuel composition and experimental setup [15]. Hence chemical kinetics also play an important role after soot precursors have been formed. The particle inception stage initiates soot production in the sense that the growth and oxidation of soot particles occur in a qualitatively different manner than the P A H forma-tion chemistry. Once the smallest soot particles are formed from the precursor P A H , soot inception takes place at molecular masses between 500 and 2000 amu, where conglomer-ation of the small molecules forms the particle-like structures. Condensation reactions of A . l . Soot Formation Mechanisms 106 the gas-phase precursor species lead to the appearance of the first recognizable soot par-ticles (also called nuclei). These soot nuclei are very small (d < 2 nm) and the formation of large numbers involves negligible soot mass in their formation regions. The detailed chemistry for the nuclei non-equilibrium formation process is highly complex, with several theories on the pyrolysis process that leads to particle nucleation. For example, thermal cracking can result in fragmentation of fuel molecules into smaller ones, condensation • reactions and polymerization that result in larger molecules, and dehydrogenation that lowers the H / C ratio of the hydrocarbons destined to become soot. Depending on the formation temperature, three different paths to soot production appear to exist. A t low temperatures (< 1700 K ) , only aromatics or highly unsaturated aliphatic compounds of high molecular weight are very effective in forming solid carbon through pyrolysis. At intermediate temperatures typical of diffusion flames (> 1800 K ) , all normally used hydrocarbon fuels produce soot if burned at sufficiently rich stoichiometry, although by following a different path. A t very high temperatures (above practical application ranges), another nucleation process involving carbon vapour is likely to occur. Figure A.2 shows proposed formation mechanisms for low and intermediate temperatures, and has consid-erable experimental support. At low temperatures, an aromatic hydrocarbon can produce soot via a relatively fast direct route that involves condensation of the aromatic rings into a graphite-like structure. After the transformation of the initial hydrocarbons into macro-, molecules, the partial pressure of these molecules grows until supersaturation becomes sufficient to force their condensation into liquid microdroplets. These become soot nuclei, while subsequently formed gaseous macromolecules then contribute to nuclei growth. Ex-periments on the pyrolysis of benzene between 1300 K and 1700 K support the physical condensation mechanism and the direct path of Figure, A.2. Above 1800 K, however, a slower and less-direct route is favoured that entails ring break-up into smaller hydrocar-bon fragments. These fragments then polymerize to form larger unsaturated molecules that ultimately produce soot nuclei. Aliphatic molecules can only follow this latter route, which is also supported by experimental flame studies where polyunsaturated hydrocar-bon compounds are involved in nucleation, and acetylenes and polyacetylenes have been detected that decrease in concentration as the mass of carbon formed increases. A . l . Soot Formation Mechanisms 107 Aromatics Condensation reactions ' i Direct (fast) Sool J ^ H , Indirect (slow) C H , Soot Aliphatics F igu re A . 2 : Particle Inception Process [15] A.1.3 Particle Growth Once soot nuclei are formed in the inception stage, they can grow by two related mechanisms: collisional coagulation (a physical process) and surface growth (a chemical process). The majority of the solid phase soot material (> 95%) is generated at this growth stage. When the particles are small (d < 10 nm), collisions between them generally lead to the formation of larger spheroids (while decreasing the total particle number) via the process of coagulation. Coagulation is essentially a sticking process of particles, which subsequently are 'glued' together by a common outer shell generated by deposition. If the soot spherules have solidified before collision and surface growth rates have diminished, the resulting particles resemble a cluster in which the original spherules retain much of their individual identity. The rate of coagulation to larger particles is very sensitive to number density. For example, the number of soot particles decreases rapidly with advancing crank angle in the diesel engine during the early part of the expansion process, and therefore the coagulation and aggregation processes are essentially complete well before the exhaust valve opens [15]. Surface growth of particles proceeds in conjunction with coagulation, and the mechanism is assumed to be similar to the formation of P A H . However, it is not a gas-phase reaction of small molecules, but a heterogeneous process where absorption and desorption processes at the surface have to be considered as well. Surface growth is essentially the gas-phase deposition of hydrocarbon intermediates (through chemical reactions) on the active surface sites of soot spherules. During coagulation when the particle are small, rapid surface growth will quickly restore the original spherical shapes of the constituent primary soot particles. A . l . Soot Formation Mechanisms 108 The major growth species in hydrocarbon flames is acetylene, although PAHs may also play a role [29]. Soot aggregation takes place in the final steps of the growth stage, when coagulation is no longer possible due to the lack of surface growth. Consequently, continued coalescence of soot particles results in open structured aggregates as shown in Fig. 2.2, containing up to thousands of spherules in a fractal-like geometry and characterized by a log-normal distribution. There are likely significant electrostatic forces on the individual particles, with the positive charges responsible for the chain-like structure. Figure A.3 shows typical variations of quantitative soot parameters during the inception and growth stages, as a function of time. If the particles are assumed to be monodisperse, fv is related to N and d by fv = (^)Nd3 (A.3) The rate of change of particle number density with time can be expressed as dN = Nn - Na (AA) where N n is the rate at which fresh soot nuclei is produced and N a is the rate of coagulation. At the peak of the N curve, there is a balance between these two rates. To the left of the peak ( N n > N a ) , the particle diameter remains essentially constant at the minimum detectable diameter and the small rise in soot volume fraction is dominated by soot inception from P A H . To the right of the N curve peak (N„ > N n ) , the number of coagulation collisions is high because of the high number density, and at the same time nucleation ends because there is enough dispersed surface area for gaseous deposition of hydrocarbon intermediates so the probability of generating new soot nuclei falls to zero. Wi th nucleation halted slightly to the right of the N curve peak, the majority of the subsequent increase in soot volume fraction comes from surface growth. The particle number density can decrease in several orders of magnitude further to the right of the N curve peak. This results from coagulation, which is responsible for partly increasing the particle diameter, but does not contribute to the increase in soot volume fraction. Surface growth that takes place'on nuclei and on spherules is responsible for forming the concentric shells that constitute the outer portions of spherules, which are distinct from the less organized spherule center. In addition, it can be seen from Figure A.3 that the A . l . Soot Formation Mechanisms 109 H / C ratio of the hydrocarbons formed in the pyrolysis and nucleation process and of the soot particles continually decreases. The H / C ratio starts from around 2 (typical of common fuels) and decreases to of order 1 in the freshest soot particles capable of being sampled, and then to about 0.2 once surface growth has ceased. In the latter stages of soot formation, the addition of mass to the soot particles occurs by reaction with gas-phase hydrocarbon (mainly acetylene). Due to the preferential addition of larger polymers, the H / C ratio decreases toward a steady-state value shown. Thus most of the polyacetylenes added must be of very high molecular weight. F i gu re A . 3 : Quantitative Soot Parameters During Inception and Growth [15] A. 1.4 Oxidation Soot oxidation is a heterogeneous reaction, the rate of which depends on the diffusion of reactants to and of products from the surface, as well as the kinetics of the reaction. For particles less than about 1 fj,m in diameter (i.e. most soot aggregates), diffusional resistance is minimal and therefore oxidation is kinetically controlled [15]. Since it is difficult to experimentally follow the oxidation of soot particles in flames, much theory remains to be understood about this process. Soot particles formed during combustion can be oxidized by O atoms, O H radicals, and O2. Other oxygenated species such as H2O and CO2 may be important under special conditions. Although O H and O radicals are- more reactive than O2, their concentrations are much less than that of oxygen. A . l . Soot Formation Mechanisms 110 Since particle oxidation is directly proportional to the available carbonaceous surface area, molecular O2 is generally considered to be the main oxidizer of soot [42]. The O H radical may be important in oxidation in the flame zone under near stoichiometric conditions. Oxidation is essentially an exterior surface phenomenon, with the surface to volume ratio being the main measure of the effect of particle geometry on oxidation rates. Therefore coagulation and aggregation will decrease oxidation rates because of the relative reduction in surface area. In premixed flames, oxidation can occur at every stage of combustion. In contrast, the bulk of oxidation in diffusion flames occurs after the particles leave the flame front, because the hydrocarbon fuel is fully decomposed in its derivative radicals in a pyrolysis zone before being oxidized. Since the equivalence ratio (EQR) varies from point to point in diffusion flames, the oxidation mechanism is linked to the local presence of oxidant, and hence to the diffusive phenomenon. Ox-idation in this case also continues downstream as long as the temperature is high enough. Although thermodynamics alone cannot adequately describe the soot formation phe-nomenon, it can dramatically affect the rates of soot inception and growth, particularly those relating to the detailed balancing of reactions [29]. For instance, one of the P A H forming elementary reactions is the reversible sequence: C6H5 + C2H2 «-» (C6H5C2H2) <-+ CeH5C2H + H (A.5) The net forward rate of this reaction is in general limited by the reversible reactions. The kinetic expression k f C e H s ] ^ ^ ] represents an upper limit for the net rate of production of both C6H5C2H2 and C 6 H 5 C 2 H . Therefore the prediction of P A H and subsequent soot formation depends significantly on the selected thermodynamics. Other parameters affecting soot production include the addition of inert gases to the fuel stream, where the effect is through the temperature, with those having greater specific heats resulting in less soot yields. In terms of fuel composition, the amounts of aromatic hydrocarbons, oxygenated compounds, and metal additives can affect soot formation [9]. An increase in aromatics will greatly increase the soot yield, while increasing the number of oxygen atoms allows the direct removal of carbon atoms as potential soot particles. Metal additives such as nickel and manganese reduce soot formation because their low ionization potentials cause the soot particles to become charged and aggregation is re-A.2. Shock Tube Apparatus 111 duccd, thus increasing the likelihood of oxidation. Certain chemical additives can promote pyrolysis rates and thus increase sooting tendencies. For example, sulphur trioxide can suppress soot in diffusion flames and increase soot in premixed flames. Another hypoth-esis for aromatic fuels is based on the soot precursor stuctures being resonance stablized at high temperatures. The rather conjugated aromatic structures are so stablized that, besides having an elemental form of the final soot structure, this resonance creates the high sooting tendency. Figures A.4 to A.6 show the detailed schematics of the injector connection, optical access section, as well as the principle used to detect light emissions from the combustion event. Table A . l summarizes previous shock tube experimental studies of in-flame and emissions measurements of particulate matter. Figure A.7 illustrates the shock tube gas and particle sampling apparatus at the University of Illinois at Chicago. A.2 Shock Tube Apparatus Injector Nozzle^ Quartz Window F i g u r e A . 4 : Injector Connection Setup [21] A.3 Previous Work • . . S u m m a r y o f In -F lame P M E x p e r i m e n t s M u i n o r \ s ; r e a r Fue l ( s ) R e a c t i o n P r e s s u r e (bar) Te m pe r a t u r e (K) M e a s u r e m e n t Q u a n t i t y M e a s u r e m e n t M e t h o d Alexiou, Williams 1995 toluene/n-heptane, toluene/iso-octane Pyrolysis 0.19-0.36 1600-2400 soot induction time, formation rate laser beam attenuation 632.8 nm Alexiou, Williams 1996 toluene/methanol, toluene/ethanol, etc. Pyrolysis 1.7-3.5 1450-2450 soot induction time, formation rate laser beam attenuation 632 8 1152 0 nm Cadman 1996 ethanol Injection 12 1000-1800 soot yield laser absorption 488 nm Cadman, Denning 1996 oxygen Premixed 4-14 1500-3500 soot oxidation rate laesr beam attenuation 488 633 nm oadman, Denning 1997 hydrogen/oxygen Premixed 3-15 1150-3000 soot oxidation rate laser absorption 488 nm Kellerer, Wittig 2000 toluene, benzene, ethylene, methane, etc. Premixed 10-60 1500-2300 soot particle size, number density, volume fraction laser light scattering 488 632 8 nm Kellerer, Muller 1996 methane, ethylene, acetylene, propane, etc. Premixed 15-100 1600-2100 soot particle size, number density, volume fraction wV^ iU M i l l laser light scattering 488 632 8 nm Kunz, Cadman 1997 propanol/tetradecane/ water Injection 4-9 1000-2200 soot yield • w . i n if laser absorption 488 nm Muller, Wittig 1991 methane, ethylene, acetylene Premixed 4-9 1600-2000 soot induction time, particle size, concentration optical dispersion quotient; 488, 633 nm Park, Appleton 1973 oxygen Premixed 0.05-13 1700-4000 soot oxidation rate laser-light transmission fi39 R nm i — \ S u m m a r y o f P M E m i s s i o n s E x p e r i m e n t s M u u i o q s j r e a r Fue l ( s ) R e a c t i o n P r e s s u r e (bar) Te m pe r a t u r e (K) M e a s u r e m e n t Q u a n t i t y M e a s u r e m e n t M e t h o d Sidhu, Graham 2001 CNG, DME, biodiesel, diesel Premixed Injection 20-27 1000-1500 particle size, shape, SOF filter, gravimetic, SEM, thermal desorption Wang, Cadman 1995 benzene Injection 2 2400 C 6 0 filter, mass Wang, Cadman 1998 toluene, n-heptane, propanol-1 Injection 2-25 1000-3000 soot yield, PAH yield spectrometry filter, gravimetic A.3. Previous Work 113 Manual Valve ' ^  ; Solenoid Valve Quartz Windows (3) Injector Optical Test Section Pressure Transducers Adaptor ^ Clamping Plates Rubber Instrumentation Window F igu re A . 5 : Optical Access Section [21] Ignition inside the shock tube LShock tube wall V .Quartz window Optical fiber Bandwidth filter Photo multiplier J \ r V Data acquisition system o o o F igu re A . 6 : Optical Access Principle [19] A.3. Previous Work 114 Sampling High Pressure Sbocktube Croup Sampling Samples are withdrawn from the high-pressure shock tube through a port in the endwall of the driven section into two sample vessels. One vessel is used to collect a sample of the reagent mixture before the shock tube is fired. A sample of the gases after the shock tube has been fired is collected in the second vessel. The duration for which is a sample is collected has been determined empirically to ensure that only gases close to the end of the shock tube are taken when acquiring post-shock samples. The rig is configured so that when very condensable species are being sampled a 'flush and sample' method can be used. Additionally, the whole sampling rig can be purged with nitrogen and evacuated with a turbo pump independently of the shock tube. The pictures on this page show the sampling rig attached to the high pressure shock tube. The large blue objects are air actuated valves that are controlled from a computer that ensures that samples are taken at the correct point in an experiment. The large silver coloured vessel close to the end of the shock tube collects the post-shock sample. The pre-shock sample vessel is much smaller and cannot be easily seen in these photographs. Two views of the sampling rig attached to the high-pressure shock tube. V\( PcuirOrirrr! of C t i c r i i K . i l I r ic j incrr incj F i g u r e A . 7 : Shock Tube Sampling at the University of Illinois at Chicago [5] Appendix B Diaphragm Material Selection and Testing B.l Summary The search for appropriate diaphragm materials has evolved over time, with each new trial material having its own advantages and drawbacks. Although some drawbacks can be resolved, Lexan plastic has been found to produce the best compromise between perfor-mance consistency and P M measurement disturbance. A l l potential diaphragm materials and thicknesses are first tested in the shock tube for burst pressures, with their repre-sentative throttling loss coefficients determined during subsequent actual experiments. Burst pressures are measured by the voltage difference on the pressure transducer, while loss coefficients are determined by comparing the calculated and actual experimental conditions. It was found that each diaphragm type produced very consistent burst pres-sures and loss coefficients, with the latter being more dependent on the material. There seems to be minimal batch to batch variations (in material properties and manufacturing defects) for all the diaphragms tested. Any discrepanies above the natural variability should be noted and investigated. B.2 Material Selection The bursting characteristics of stainless and carbon steel shims, as well as Lexan di-aphragms are shown in Figures B.2 and B.3. Figure B.4 gives an example of pressure 115 B.3. Burst Pressure Testing 116 spikes caused by flying diaphragm pieces. Figure B.5 shows several versions of trap de-vices, in an effort to eliminate the pressure disturbances. F i g u r e B . l : Stainless Steel Shim Diaphragm F i g u r e B .2 : Carbon Steel Shim Diaphragm B.3 Burst Pressure Testing To determine the burst pressure of a particular diaphragm type, one or two (depending on the bursting consistency) diaphragms from each new batch are tested individually B.4. Determination of Throttling Losses 117 Figure B.3: Lexan Diaphragm in the shock tube. The Eclipse high-pressure transducer is initially used to measure the voltage at atmospheric pressure (V 0 ) . After closing and fastening the shock tube, the driver section pressure is then slowly increased until the diaphragm bursts, at the burst pressure voltage. The following experimentally determined [19] conversion factor is used for converting measured voltages to pressures. Table B . l shows a summary of the diaphragms tested for burst pressure. P(psi) = 1250 x (Vtont - VQ) (B.l) B.4 Determination of Throttling Losses When the diaphragm bursts, throttling losses will occur due to small pieces that remain attached to the clamped portion, and the resultant decrease in flow area (see Figure B.6). Since the variations in driver gas properties due to throttling can affect the level of tailor-ing, empirical loss coefficients must be determined for each diaphragm type to describe its burst performance. The method of compensation involves using a polytropic ratio to obtain an actual driver pressure that is higher than the theoretically calculated P4. Assuming the driver gas expansion process behaves as an ideal gas undergoing polytropic expansion, the polytropic coefficient (n) is governed by B.4. Determination of Throttling Losses 118 3.5 2.5 1.5 0.5 . L 1 1 F ^ W^. ^ W u , i . i _ . Time (ms) 15 F igu re B .4 : Pressure Disturbances Within Experimental Region (40 bar) n — (polytropic ratio) x k (B.2) PVn = constant = P{V? = P2Vf (B.3) By decreasing the polytropic ratio from the isentropic value of one, the expansion pro-cess deviates from isentropic and results in irreversible losses [46]. During the first few experiments with every new diaphragm type, the best representative polytropic ratio is continually estimated by comparing the predicted and measured experimental conditions. The ratio that produces the closest agreement between the conditions will be used for all subsequent experiments with that diaphragm type. Table B.2 gives the polytropic ratios suitable to each diaphragm type, based on the applicable experiments. B.4. Determination of Throttling Losses "•BBH_H_B__B_MM__H__M F i g u r e B . 5 : Diaphragm Fragment Trapping Devices Diaphragm Flow separation and re-circulation zone causing throttling loss F igure B .6 : Throttling Losses [19] B.4. Determination of Throttling Losses 120 Tab le B . l : Summary of Burst Pressure Testing Material Thickness Transducer Voltage (V) Burst Pressure Atmospheric Burst (psi) Brass Shim 0.003" 0.451 0.526 94 0.451 0.522 89 Stainless Steel Shim 0.001" 0.448 0.542 118 Stainless Steel Shim 0.002" 0.447 0.63 229 . Stainless Steel Shim 0.003" 450 * Stainless Steel Shim 0.004" 0.451 1.129 848 Carbon Steel Shim 0.002" 0.441 0.507 83 Carbon Steel Shim 0.003" 0.448 0.549 126 Carbon Steel Shim 0.004" 0.448 0.587 174 Carbon Steel Shim 0.006" 0.45 0.65 250 Carbon Steel Shim 0.007" 0.452 0.768 395 Carbon Steel Shim 0.008" 450* Transparency Film 0.003" 0.451 0.551 125 Lexan (matte) 0.010" 0.451 0.53 99 Lexan (clear) 0.010" 0.447 0.549 128 0.447 0.549 128 Lexan (clear) 0.020" 0.447 0.635 235 0.447 0.644 246 Lexan (clear) 0.030" 0.451 0.765 393 0.451 0.739 360 0.451 0.747 370 * Note: estimated from pressure regulator gauge Tab le B .2 : Summary of Diaphragm Polytropic Ratios Material Thickness Polytropic Ratio Brass Shim 0.003" 0.95 Stainless Steel Shim 0.001" 0.95 Stainless Steel Shim 0.002" 0.95 Stainless Steel Shim 0.003" 0.95 Stainless Steel Shim 0.004" 0.95 Carbon Steel Shim 0.002" 0.95 Carbon Steel Shim 0.003" 0.95 Carbon Steel Shim 0.004" 0.95 Carbon Steel Shim 0.006" 0.95 Carbon Steel Shim 0.007" 0.95 Carbon Steel Shim 0.008" 0.95 Transparency Film 0.003" 0.95 Lexan (matte) 0.010" 0.9 Lexan (clear) 0.010" 0.9 Lexan (clear) 0.020" 0.9 Lexan (clear) 0.030" 0.9 Appendix C Contamination Control To aid in the control of all possible contamination sources, various diagnostic tools and dedicated instruments are used to detect and analyze particle and gas contamination forms, with the relevant figures and graphs summarized below. C . l Contamination Detection F igu re C . l : Passing Through Shock Tube 121 C.2. Pre-Experiment Control 122 4 3 5 -(ng/min) 3 -(ng/min) 2 5 -ncrement 2 i ncrement 1 5 -> ass 1 1 -_> 0.5 - • • < 0 -C i | 1 1 2 4 6 8 Time (minutes) 10 Figure C . 2 : Bypassing Shock Tube C.2 Pre-Experiment Control Table C . l : Summary of Total Particle. Counts Particle Count (#/ccm) Ambient Lab Air Filtered Bottled Air, Through Tube Filtered Bottled Air, Bypassing Tube > 50000 - 5 0 ~ 2 C.3 Post-Experiment Control When a gas jet carrying particles impinges upon a surface, those larger than a critical size will impact .on the surface, while the rest will follow the streamlines and remain airborne. The main parameter controlling inertial impactors is the Stokes number (St) at the impaction conditions, which depends on the particle mass, flow properties, and the interactions between the particle and the host fluid. For hypersonic inertial impactors, the Stokes number is expressed as [36] St = 32.02(5o —) (C.l) C .3. Post-Experiment Control 0.1, 0.09 0.08 0.07 r-I 0.06 TS 0.05 c g> w 0.04 Z gi - 1 0.03 0.02 0.01 0 Time (ms) Figure C.3: Blank Test Light Emission (20 bar, 1200 K ) 0.01 0.009 h 0.008 0.007 J> 0.006 o % 0.005 0.004 0.003 0.002 0.001 10 15 Time (ms) Figure C.4: Inert Blank Test Light Emission (20 bar, 1200 K) C.3. Post-Experiment Control 124 Contaminated Clean Figure C . 5 : Comparison of Tube Surfaces Figure C .6: Towel and Brush Cleaning Figure C.7: High-Speed Rotating Brush Cleaning Setup [41] C.3. Post-Experiment Control 125 Figure C . 9 : Magnetic Filter C.3. Post-Experiment Control 126 where L is the nozzle to collector plate distance, dn is the nozzle diameter (as shown in Figure C.10), and S0 is the particle Stokes number at stagnation conditions upstream of the nozzle. After substituting the expression for S0 and assuming m p ~ d a 3 (C.2) where mp is the particle mass and da is the aerodynamic diameter. The linear expression of St with the particle diameter is st = 32m0M8?f';d'c°7"r) (c.8). where P0 is the gas stagnation pressure, pp is the particle density, ca is the carrier gas sound speed at the stagnation temperature, and the mass and aerodynamic diameters are assumed to be the same [36]. Thus, for given particle properties and stagnation conditions, the particle Stokes number varies linearly with ^ , which can be used to control the cutoff diameter. For large . values of ^ , StS>l and the particle accelerates enough during the jet expansion to impact against the collector plate. As jfc gets smaller, the particle maximum velocity reduces until the critical Stokes number is reached, where the velocity is insufficient to contact the plate and the particle will remain airborne. The theoretical collection ef-ficiency as a function of ^ exhibits a step function resemblance for hypersonic impactors. Po dn i p 1 / L I < Figure C .10: Schematic of Hypersonic Impactor C.4. Particle Identification Figure C . l l : Impactor C . 4 Particle Identification C.4.1 SEM Analysis CA. Particle Identification 128 impactor # of holes d„ L 1 3 1/4" 1/8" 2 5 3/8" 3/16" 3 10 5/8" 5/16" impaction bleed flow F i g u r e C .12 : Multi-Stage Impactor Schematic F i g u r e C .13 : Single Stage Multi-Nozzle Impactor: Nozzle and Impaction Plates CA. Particle Identification Figure C.14: Dust Particle kCts 15H 104 Element Concentration Oxygen 47.64 wt% Sodium 10.84 wt% Magnesium 1.49 wt% Aluminum 1.49 wt% Silicon 26.00 wt% Potassium 0.47 wt% Calcium 3.71 wt% Iron 8.12 wt% Molybdenum 0.23 wt% 54 Na AA t ? -i 1—i r 8 keV Figure C.15: Dust Composition C.4. Particle Identification Tab le C . 2 : Elemental Composition of Various Materials Material Elemental Composition (ADDTOX %) Soot (Black Carbon) C, H Carbon Steel Fe, C, Mn Stainless Steel Fe, C, Cr, Ni, Mn, Si Ambient Dust Mg, Al, Si, S, Ca, Fe Towel Fiber (Cotton) C, 0, H Nylon C, 0, N, H Rubber C (88), H (12) Teflon F (76), C (24) Lexan (polycarbonate) C (76), 0(18), H (5), Si TEM main grid Cu TEM laces C, 0, H SEM sample holder C C.4.2 T E M Analysis to Aethalometer inlet TEM grid holder F igu re C .16 : T E M Sampling Setup C.4. Particle Identification 131 Figure C .18: Lexan Particle (100 nm) C .4 . Particle Identification 132 F igu re C.20: Lexan Particle (150 nm) C.4. Particle Identification Counts -f 200 -J F igu re C.22: Soot Particle 2 CA. Particle Identification F igu re C .23 : Soot Particle 3 F igu re C .24 : Soot Composition CA. Particle Identification 135 F igu re C.26 : T E M Soot Photograph 2 C.4. Particle Identification 136 Table C.3: S E M Analysis of Lacy Carbon Grid from Blank Experiment Elemental Composition (Approx. %) Size (um) Shape Material Background 1 Background 2 Particle 1 Particle 2 Particle 3 Particle 4 Particle 5 Particle 6 Particle 7 Particle 8 Particle 9 Particle 10 Particle 11 Particle 12 C, O (44), Cu (45), Si (10) C, O (56), Cu (43) C, O, Cu, Si C, O, Cu, Si C, O, Si C, 0 C, O (44), Cu, Si (7) C, 0(37), Cu (50), Si (13) C, O (45), Cu (45), Si (8) C, O (61), Si (2), P (9), Ca (18), Mg (0.5) C, O (14), Cu (2), Si (83) 0(9), Al (1), Si, P, CI, K(87), Ca . C, O, Cu, Mg, Al, Si, Ti C,0 (60), Si (25), Na, Cu (11), Al 1 oval Lexan 1 oval Lexan 0.5 rectangular 1 needle 0.5 circular Lexan 1.5 Lexan 0.5 Lexan 2 dust 2 Lexan 3 8 dust 0.5 Particle Density ~ 20-30 particles per square grid (40 um X 40 um) Tab le C . 4 : S E M Analysis of Lacy Carbon Grid from Sooty Experiment . Elemental Composition (Approx. %) Size (um) Material Particle 1 C, O (39), Cu (58), CI 8 Particle 2 C, 0(51), Cu (36), Si (9), CI 1 Lexan Particle 3 C, CI 5 soot Particle 4 C, O (38), Cu (58), CI (3) 0.5 Lexan Particle 5 C, 0(39), Al (11), Cu(48) 1 Particle 6 C, O (55), Cu (44) 1 Particle 7 C 4 soot Particle Density ~ 60-80 particles per square grid (40 um X 40 um) C.4. Particle Identification Tab le C . 5 : T E M Analysis of Lacy Carbon Grid from Blank Experiment Elemental Composition Size (nm) Material Background 1 Cu Laces Background 2 Cu Laces Particle 1 C, Cu 100 Soot Particle 2 C, 0, Si, Cu 100 Lexan Particle 3 C, 0, Si, Cu 100 Lexan Particle 4 C, 0, Si, Cu 100 Lexan Particle 5 C, Cu 100 Soot Particle 6 C, 0, Si, Cu 50 Lexan Particle 7 C, 0, Si, Cu 100 Lexan Particle 8 C, 0, Si, Cu 100 Lexan Particle 9 C, 0, Si, Fe, Cu 100 Particle 10 C, O, Si, Cu 100 Lexan Particle 11 C, 0, Si, Cu 100 Lexan Tab le C . 6 : T E M Analysis of Lacy Carbon Grid from Sooty Experiment Elemental Composition Size (nm) Material Particle 1 C, O, Si, Cu 50 Lexan Particle 2 C, O, Si, Cu 50 Lexan Particle 3 C, O, Si, Cu 100 Lexan Particle 4 C, Cu 400 Soot Particle 5 C, Si, Cu 50 Particle 6 C, Cu 500 Soot Particle 7 C, Cu 400 Soot Appendix D Particle Loss Minimization D . l Inside Shock Tube Calculations of van der Waals adhesive forces between soot particles and shock tube walls are shown below. F a d h = (D.l) Ad 12x where A is a material-dependent constant, d is the particle diameter, and x is the separa-tion distance (based on the surface roughness). Typical shock tube relevant parameters result in (75 x 1CT 2 0 J) x (500 x 10- 9 m) , , F ™ * = - 12(200 x 1 0 - m)' L , x ^ * ( D ' 2 ) Estimates of thermophoretic force and velocity of typical soot particles produced during combustion are shown below. Fth = -=—= (D.3) -p\d?VT ~T where p is the gas pressure, A is the gas mean free path, V T is the temperature gradient, and T is the temperature of the particle. -0 .55r /VT , ^ Pg1 where rj and pg are the gas viscosity and density, respectively. 138 D.2. Gas Venting Process 139 Using a representative V T of 30000 K / m (as it varies with the radial distance from the tube center) as well as gas properties based on typical experimental air/helium propor-tions, Fth and Vth values are estimated as - (1 x 10 5 Pa)(0.16 x 10~ 6 m)(500 x 1 0 ' 9 m)2(30000 K/m) „ Fth = — ^ '-> ~ - 4 x 1 0 - TV (D.5) -0.55(198 x 10- 7 /Vs/m 2)(30000 K/m) n n , / r v . (0.262 k,/J)WK) ' ~ 0-0042 m/ , (D.6) Settling velocities of.sample soot particles are shown below. where pp is the particle density and g is the acceleration due to gravity. V - ( 2 5 Q 0 ^ / m 3 ) ( 5 0 X 1 0 " 9 m ) 2 ( 9 ' 8 0*m-TmU (TSR\ V t S ~ 18(198 x 1 0 - ' Ns/m>) 2 X l ° ™/S ( D " 8 ) V - ( 2 5 ° ° W m 3 ) ( l x l 0 - « m) 2(9.8 m/s 2 ) 7 v i n _ 5 « m J V t S ~ 18(198 x 1 0 - ' Na/m') 7 X 1 0 m / s ( D ' 9 ) The diffusion coefficients for sample 50 nm and 500 nm diameter soot particles are cal-culated below. where k is Boltzmann's constant and C c is the slip correction factor (dependent on particle size). ' (1.38 x 1 0 - 2 3 J/K){300 A")(1.3) r o 2 , , n . " _ (1.38 x 1Q- 2 3 J/AT) (300 AT) (5.0) q , ^ ~ 3,(198 x 10- 7 iVs/m 2 ) (50 x 10~* nm) ~ 2 " 2 X 1 0 m ' s (D-12) D.2. Gas Venting Process 140 Tab le D . l : Summary of Gas Venting Durations Initial Pressure Final Pressure Average Duration (bar) (bar) (s) 3 1 4.90 5 1 7.19 8 1 8.85 . 1 0 1 9.64 12 1 10.50 14 1 11.48 22 1 16.25 * * Note: obtained using extrapolation D.2 Gas Venting Process D.3 Sample Container The electrostatic force between two charges are given by Coulomb's law: FE = K B ^ (D.13) where q and q' represent the charge magnitudes, R is the separation distance, and KB is a proportionality constant. The terminal electrostatic velocity can be obtained by equating the force to the Stokes drag, and is given by (for particle motion in the Stokes region): VTF, 2>Trr)d (D.14) where e is the charge of an electron, and the particle having n elementary units of charge in an electric field E. 1 Appendix E Aethalometer Data Analysis E . l Operation Principle The optical attenuation (ATN) value is defined with respective to the transmitted light intensities through two different portions of the filter spot. ATN = -100 x ln( ! ~ * ) (E. l ) In general, the optical intensity functions are products of wavelength-independent terms. The intensity of light detected.after passing through a blank portion of the filter is 70(A) = 71(A) x F(X) x OC(X) x D(\) (E.2) where IL(A) is the emission intensity of the light source, F(A) is the spectral transmission function through the filter, OC(A) is the spectral transmission function through all the other optical components, and D(A) is the spectral response function of the detector. When the same light source and detector is used to measure the optical transmission through an aerosol deposit on the filter, the net intensity will be / .= 70(A) x e~AW (E.3) . where the absorbance due to a particular aerosol species is A(X) = k(\) x MP (E.4) A and Mp is the amount of the particle species whose optical absorption is inversely 141 E . l . Operation Principle 142 proportional to the wavelength A. For black carbon measured with white light in Equation E . l , an A T N value of 1 is barely perceptible (a contrast between deposit and blank of only 1%), while 100 corresponds to an aerosol spot that is dark gray (approximate carbon loading of 6 /zg/cm 2). This measurement is affected by the wavelength of the light, as the absorption of light by a broad band absorber such as graphitic carbon is inversely proportional to the wavelength of the light used. For a given mass of black carbon (MBC), the optical attenuation at a fixed wavelength A can be expressed as ' ATN(\) = (a*)(\) x MBC .(E.5) where (cr*)(j) is the optical absorption cross-section that is wavelength dependent (i.e. not a physical constant), and called the specific attenuation' (CT). Since the absorption spectrum will vary between different aerosol species, the specific attenuation value must be determined for each wavelength used by the Aethalometer (see Table E . l ) , in order to correctly convert the A T N measurements to various species mass results. The wavelength dependence of optical components and filter configuration is weighted by the j function. Tab le E . l : Summary of Specific Attenuation Values Channel Wavelenqth °" (Black Carbon) cr (Elemental Carbon) Infrared (IR) 880 nm Ultraviolet (UV) 370 nm 16.6 12.6 39 5 30.0 Intercomparisons with other analytical techniques were used to validate the Aethalome-ter's optical attenuation method [14] [17] [35], since black carbon measurements are highly method and instrument dependent. Very good agreements with techniques such as Particle Soot Absorption Photometer, Laser Integrating Plate, T O R Thermal Oxi-dation, and R & P Carbon Analyzer were found. Elemental carbon is a commonly used alternative definition from the filter-based T O R thermal method, where the black car-bon is interpreted as a subset of the elemental carbon mass. The second channel on the Aethalometer illuminates in the near-ultraviolet (UV), where certain organic compounds (e.g. P A H , fresh diesel exhaust) show strong spectrally-specific absorption (as compared to the broad spectrum absorption of black carbon). Numerically, the U V channel beam will result absorbance of E . l . Operation Principle 143 A(A*) = fc(l)xAf/» + ^ ( P ( A * ) x C ( P ) ) . (E.6) where P(A*) is the U V absorbance (at the shorter wavelength A* of the quantity C of compound P, summed over all relevant compounds, each with a different absorption effciency P at each different wavelength A*. Since the U V absorption efficiency is highly variable, the U V beam measurements cannot be directly compared to the B C beam results to find the mass of a given species. Therefore, a fictional U V P M material is defined as if it absorbed U V photons with the same efficiency as black carbon at the U V wavelength, and is expressed in units of BC-equivalent mass. This equivalence definition means the additional amount (above the B C mass) reported by the U V channel is not a physical mass, but rather the equivalent mass of B C required to cause this additional absorption. In summary, the 880 nm wavelength black carbon measurement yields an absorbance of A(\) = k(\) x MP (E.7) A while the 370 nm wavelength U V channel measurement yields an absorbance of A(X*) = k(^) x (MBC + MUVPM) (E.8) Furthermore, the above optical attenuation derivation also assumes that the optical absorption is linearly proportional to the mass of absorbing material, through the a factor. This is found (in practice) to be true under several conditions [14], which are met in this intended P M sampling application: 1) the particle sizes are considerably smaller than the wavelength size parameter (2nA); 2) the amount of absorbing material in the sample does not lead to saturation; and 3) the effect of the embedment of the aerosol particles in a deep matrix of optically-scattering fibers is to eliminate any reduction of the optical transmission through the filter by particle optical scattering, therefore rendering the measurments sensitive only to absorption. For example, the second condition is met by advancing the tape to a clean spot when it reaches a certain density or 'blackness'. Sample calculations using the data in Table E.2 are shown below. The initial attenuation value (ATNrj) is first calculated by the instrument during tape spot calibration and stored internally. The measured attenuation during the first minute of this data set is calculated by Equation E . l and exceeds the initialization value by 1.527 (not shown in Table). Since only relative values of A T N are important, the additional A T N increase from the black E.2. Data Algorithm Modifications 144 carbon loading during the second minute is 0.046 (i.e. 3.759 - 3.713). This is obtained from the theoretical formulation where all four signal voltages are taken into account between the first and second minutes. ATN2 - ATN, = (-100 x (In 3-0161 - 0-0196 _ ( 3.0213 - 0.0196 = V V 3.2907 - 0.0191 ; ; ^ k 3.2948 - 0 . 0 1 9 1 j j (E.9) The slight difference is due to roundoff errors in intermediate values, with the datafile's A T N column containing the most accurate values. The black carbon loading and mass during the first timebase interval are ATN2 — ATNi 3 .759- 3.713 „ „ „ „ „ „ . 2. , d(BC) = - = = 0.00277 g m2 (E.10) a 16.6 mz/g MBC = d(BC) x Aspot = 0.00277 g/m2 x (0.5 cm 2 ) x (1 x 10~ 4 m2/cm2) = 1.385 x 10" 7 g ( E . l l ) [BC] — — - — ~ B ^ T T T o ^ \ = /f85 * 1 0 ~ 7 g x (looo L / m 3 ) x ( 1 Q 9 / } x Q Q l (Flow Rate)(Time) (4 L/mm)(1 mm) v / ^ v = 346.38 ng/m? (E.12) The resultant B C concentration (accounting for the A T N factor of 100) also deviates slightly from the displayed value during the second minute (343 ng/m 3 ) due to roundoff errors. E.2 Data Algorithm Modifications Table E.4 shows the processed Aethalometer data using the modified algorithm, where only the A T N column is used. For example, the mass increment during the first minute i s ATN2 - ATN, A MBC = x Aspot (E.13) E.2. Data Algorithm Modifications 145 2 602 — 2 319 MBC = 1 6 6 m 2 / g x ( ° - 5 c m 2 ) x ( 1 0 _ 4 m V c m 2 ) x (109 ng/g) x 0.01 = 8.524 ng (E.14) Table E.5 and Figure E . l show evidence of exponential decay in a typical B C mass increment curve, where the non-conductive bag is allowed to settle for one hour before sampling the second half of its contents. Starting with the third minute as a reference, the decay coefficient using various time delays are calculated to be very similar. Therefore it is used as the default curve fit to the Aethalometer data, and the 'true' initial mass increment is used to calculate the total mass. MBC = MBCo x e~kt . (E.15) l n ( 1 5 7 1 0 8 ) l n ( 1 2 6 4 1 6 ) l n f 1 1 4 6 0 8 ) ln( 1 5 1 5 1 ) L, _ X U V 184.187'' _ " V 184.187 > _ 1 1 1 V 184.187/ _ '"V 184.187 > n flQQ CW TR\ * " - 5 ~ - 1 0 ~ - 1 5 " - 7 5 0 0 3 3 ( K 1 6 ) Although a conductive bag surface minimizes the exponential decay rate, particle losses will still be present in other forms and should be accouted for by empirically fitting the mass deposition curve in each case. Figures E.2-E.5 show examples of other curve char-acteristics encountered, as well as the empirical trendlines used to fit the data. Their differences can be attributed to combinations of particle dynamics such as losses, en-trainment from surfaces, contamination, etc. For example, Figure E.2 shows relatively scattered data which is likely due to the instrument noise at low B C levels. Figures E.3 and E.4 show examples of polynomial curve fits where the B C values slowly increase before a sharp decline. This effect could be due to encountering a critical bag volume-to-area ratio where significant losses of the dominant size ranges start to occur. Finally, Figure E.5 shows the typical B C curve shape using the surface conductive bag, where a steady mass increment is achieved after the initial adjustment period. The most repre-sentative mass increment value all cases is the first steady data point. However, certain particle dynamics (and losses) can still be present in a conductive environment and hence deviations from the ideal flat curve are also treated with empirical curve fits in the same manner as above. Table E.6 and Figure 3.17 show samples of simultaneous B C and U V channel data. E.3. Instrument Maintenance 146 250 20 40 60 80 100 Time (minutes) F igu re E . l : Exponential Decay Test Curve F i t E.3 Instrument Maintenance Due to the high sensitivity of the measurement, periodic maintenance is needed to mini-mize additional sources of optical signal interference. These automated procedures include monitoring the reference beam, sensing beam, and flow meter voltages for consistency. Disassembly and cleaning of the sampling, cylinder head [14] is needed after sampling large amounts of P M (e.g. high soot-producing experiments). The glass surfaces on the cylinder head where the light beams pass through should be kept free of foreign material, such as flakes of quartz filter fibers. The optical test strip provided by the manufacturer can also be used to monitor the variability of the optics under various obstructions in the beam path. Natural variability in the reference and sensing beam signals should be monitored from the output log file. E.3. Instrument Maintenance 20 30 Time (minutes) 40 F igu re E . 2 : Linear Curve Fi t F i gu re E .3 : Polynomial Curve Fit (Type 1) E.3. Instrument Maintenance 148 Tab le E .2 : Sample Black Carbon Channel Data Time [BC] (ng/mJ) Flow Rate (LPM) Sensing Offset (d) Sensing Signal (I) Reference Offset (d0) Reference Signal (l0) ATN 11:42 365 4 0.0196 3.0213 0.0191 3.2948 3.713. 11:43 343 4 0.0196 3.0161 0.0191 3.2907 3.759 11:44 348 4 0.0196 3.0168 0.0191 3.293 3.806 11:45 -10795 3.9 0.0196 3.0625 0.0191 3.2962 2.392 11:46 -2292 3.6 0.0196 3.0696 0.0191 3.2947 2.112 11:47 1689 3.6 0.0196 3.0627 0.0191 3.294 2.319 11:48 2317 3.6 0.0196 3.0565 0.0191 3.2966 2.602 11:49 2556 3.6 0.0196 3.04/4 0.0191 3.297 2.914 11:50 2660 3.6 0.0196 3.0334 0.0191 3.2925 3.239 11:51 2645 3.6 0.0196 3.0224 0.0191 3.2911 3.562 11:52 2663 3.6 0.0196 3.0148 0.0191 3.2934 3.887 11:53 2777 3.6 0.0196 3.0051 0.0191 3.294 4.226 11:54 2782 , 3.6 0.0196 2.991 0.0191 3.2895 4.565 11:55 2790 3.6 0.0196 2.9801 0.0191 3.2886 4.906 11:56 2601 3.6 0.0196 2.9728 0.0191 3.291 5.224 11:57 2768 3.6 0.0196 2.9632 0.0191 3.2914 5.561 11:58 2742 3.6 0.0196 2.9499 0.0191 3.2874 5.895 11:59 2697 3.6 0.0196 2.9391 0.0191 3.2862 6.224 12:00 2608 3.6 0.0196 2.9324 0.0191 3.289 6.542 12:01 2698 3.6 0.0196 2.9231 0.0191 3.2893 6.87 12:02 2651 3.6 0.0196 2.9102 0.0191 3.2853 7.192 12:03 2610 3.6 0.0196 2.8998 0.0191 3.2839 7.51 12:04 2460 3.6 0.0196 2.8936 0.0191 3.2866 7.808 12:05 2495 3.6 0.0196 2.8852 0.0191 3.2869 8.111 12:06 2523 3.6 0.0196 2.873 0.0191 3.283 8.418 12:07 2506 3.6 0.0196 2.8631 0.0191 3.2816 8.722 12:08 2369 3.6 0.0196 2.8573 0.0191 3.2843 9.009 12:09 2361 3.6 0.0196 2.8495 0.0191 3.2847 9.295 12:10 2399 3.6 0.0196 2.8379 0.0191 3.2808 9.586 12:11 2272 3.6 0.0196 2.829 .0.0191 3.2794 9.861 12:12 2171 3.6 0.0196 2.8239 0.0191 3.2821 10.124 12:13 2260 3.6 0.0196 2.8169 0.0191 3.2828 10.398 12:14 2222 . 3.6 0.0196 2.8057 0.0191 3.2785 10.666 12:15 15259 4 0.0196 2.7444 0.0191 3.2721 12.693 12:16 941 3.9 0.0196 2.7427 0.0191 3.2741 12.818 12:17 518 3.9 0.0196 2.7416 0.0191 3.275 12.887 12:18 424 3.9 0.0196 2.7371 0.0191 3.2715 12.943 1 E.3. Instrument Maintenance 149 Tab le E .3 : Attenuation Value Changes Due to Gas Composition Start of Sampling (Air to Helium) Time (min.) ATN BC (ng/ma) 1 3.713 365 2 3.759 343 3 3.806 348 4 2.392 -10795 5 2.112 -2292 6 2.319 1689 7 2.602 2317 8 2.914 2556 End of Sampling (Helium to Air) Time (min.) ATN BC (ng/m3) 1 10.124 2171 2 10.398 2260 3 10.666 2222 12.693 15259 5 12.818 941 6 12.887 518 7 12.943 424 8 13.084 429 30 Time (minutes) F igu re E.4 : Polynomial Curve Fi t (Type 2) E.3. Instrument Maintenance 150 Tab le E.4: Sample Post-Processed Aethalometer. Data Time (minutes) [BC] (ng/ma) ATN Mass Increment (ng/min) 1 1689 2.319 2 2317 2.602 8.524096386 3 2556 2.914 9.397590361 4 2660 3.239 9.789156627 5 2645 3.562 9.728915663 6 2663 3.887 9.789156627 7 2777 4.226 10.21084337 8 2782 4.565 10.21084337 9 2790 4.906 10.27108434 10 2601 5.224 9.578313253 11 2768 5.561 10.15060241 12 2742 5.895 10.06024096 13 2697 6.224 9.909638554 14 2608 6.542 9.578313253 15 2698 6.87 9.879518072 16 2651 7.192 9.698795181 17 2610 7.51 9.578313253 18 2460 7.808 8.975903614 19 2495 8.111 9.126506024 20 2523 8.418 9:246987952 21 2506 8.722 9.156626506 22 2369 9.009 8.644578313 23 2361 9.295 8.614457831 24 2399 9.586 8.765060241 25 2272 9.861 8.28313253 26 2171 10.124 7.921686747 27 2260 10.398 8.253012048 28 2222 10.666 8.072289157 E.3. Instrument Maintenance Table E .5: Exponential Decay Test Data Time [BC] (ng/mJ) ATN Mass Increment (minutes) (ng/min) 1 54216 15.894 2 52277 22.303 193.042 3 49891 28.418 184.187 4 47517 34.241 175.392 5 46657 39.958 172.199 6 44251 45.378 163.253 7 43439 50.697 160.211 8 42620 55.913 157.108 9 40897 60.918 150.753 10 39640 65.768 146.084 11 35195 70.072 129.639 12 35228 74.38 129.759 13 34320 78.577 126.416 14 33681 82.694 124.006 15 32462 86.66 119.458 16 31920 90.558 117.410 17 31369 94.389 115.392 18 31170 98.194 114.608 75 7090 6.431 26.024 76 5997 7.161 21.988 77 4742 7.74 17.440 78 4109 8.243 15.151 79 3319 8.651 12.289 80 2954 9.018 1 11.054 81 2290 9.306 8.675 82 1553 9.505 5.994 83 1074 9.645 4:217 E.3. Instrument Maintenance 152 18 16 I 1 4 j? 12 1 10 E 2! o c tn in co » , 10 15 20 25 30 35 40 Time (minutes) F igu re E .5 : Steady Mass Increment Curve 250 10 Time (minutes) 15 20 F igu re E .6 : Sooty Experiment B C Mass Increment Curve E.3. Instrument Maintenance 153 Table [BC] E.6: Sample Post-Processed B C and U V Channel Data Time (minutes) BC ATN BC Mass Increment <n9 / m J) (ng/min) [UVPM] (ng/ma) UV ATN UVPM Mass Increment (ng/min) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 282 528. 629 756 778 768 811 837 820 718 830 790 720 783 754 867 832 1340 854 911 858 863 931 924 803 229 792 839 772 765 728 775 637 793 777 904 774 813 796 820 768 817 858 849 795 869 , 847 917 • 767 1.987 2.052 2.13 2.223 2.319 2.414 2.514 2.617 2.718 2.807 2.909 3.006 3.095 3.191 3.284 3.391 3.493 3.658 3.763 3.875 3.98 4.086 4.2 4.314 4.412 4.44 4.538 4.64 4.735 4.829 4.918 5.013 5.091 5.188 5.283 5.393 5.487 5.587 5.684 5.784 5.878 5.977 6.082 6.186 6.283 6.388 6.491 6.603 6.697 1.957831325 2.34939759 2.801204819 2.891566265 2.861445783 3.012048193 3.102409639 3.042168675 2.680722892 3.072289157 2.921686747 2.680722892 2.891566265 2.801204819 3.222891566 3.072289157 4.969879518 3.162650602 3.373493976 3.162650602 3.192771084 3.43373494 3.43373494 2.951807229 0.843373494 2.951807229 3.072289157 2.861445783 2.831325301 2.680722892 2.861445783 2.34939759 2.921686747 2.861445783 3.313253012 2.831325301 3.012048193 2.921686747 3.012048193 2.831325301 2.981927711 3.162650602 3.13253012 2.921686747 3.162650602 3.102409639 3.373493976 2.831325301 5645 5541 5547 5473 5415 5291 5250 5199 5190 5051 4984 4926 4821 4857 4764 4739 4989 4504 4567 4485 4416 4471 .4355 4309 4077 4195 4129 4135 4010. 4019 3996 3930 3915 3918 3919 3840 3832 3842 3771 3721 3660 3768 3661 3583 3523 3639 3518 3401 3442 15.385 17.013 18.64 20.247 21.837 23.387 24.927 26.452 27.974 29.454 30.914 32.357 . 33.769 35.192 36.586 37.973 39.432 40.749 42.085 43.395 44.685 45.991 47.262 48.52 49.709 50.933 52.137 53.343 54.512 55.683 56.847 57.993 59.133 60.274 61.411 62.524 63.636 64.752 65.846 66.927 67.989 69.081 70.144 71.184 72.207 73.261 74.279 75.264 76.262 20.60759494 20.59493671 20.34177215 20.12658228 19.62025316 19.49367089 19.30379747 19.26582278 18.73417722 18.48101266 18.26582278 17.87341772 18.01265823 17.64556962 17.55696203 18.46835443 16.67088608 16.91139241 16.58227848 16.32911392 16.53164557 16.08860759 15.92405063 15.05063291 15.49367089 15.24050633 15.26582278 14.79746835 14.82278481 14.73417722 14.50632911 14.43037975 14.44303797 14.39240506 14.08860759 14.07594937 14.12658228 13.84810127 13.6835443 13.44303797 13.82278481 13.4556962 13.16455696 12.94936709 13.34177215 12.88607595 . 12.46835443 12.63291139 E.3. Instrument Maintenance Tab le E.7: Sooty Experiment B C Data Time [BC] ATN Mass Increment (minutes) (ng/m3) (ng/min) • 1 18152 9.253 2 54216 15.894 200.0301205 3 52277 22.303 193.0421687 4 49891 28.418 184.186747 5 47517 34.241 175.3915663 6 46657 39.958 172.1987952 7 44251 45.378 163.253012 8 43439 50.697 160.2108434 9 42620 55.913 157.1084337 10 40897 60.918 150.753012 11 39640 65.768 146.0843373 12 35195 70.072 129.6385542 13 35228 74.38 129.7590361 14 34320 78.577 126.4156627 15 33681 82.694 124.0060241 16 32462 86.66 119.4578313 17 31920 90.558 117.4096386 18 31369 94.389 115.3915663 19 31170 98.194 114.6084337 20 30280 101.89 111.3253012 Appendix F Aethalometer Flow Meter Correction and Calibration F . l Flow Rate Correction The Aethalometer uses a thermal mass flow meter (Sierra Instruments 824 Top-Trak) to determine the volume flow rate and subsequently the black carbon concentration. It is a heated tube style device where the gas is heated as it flows through a precision sensing tube [22]. A bridge circuit measures the temperature differential between the upstream and downstream sensors, which is directly proportional to the number of gas molecules or mass flow. The original factory calibration is performed using air, at 0 and 5 L P M (at standard conditions). The volumetric flow rate is then calculated by the internal algorithm, based on the calibration gas properties. However, due to the significant amount of helium present in each shock tube experiment, the apparent sensed volumetric flow rate must be corrected for the relatively lower density and higher specifc heat of helium. This correction is performed using two different methods, a direct correction and a manufacturer-suggested procedure. To enable meaningful comparisons, a typical shock tube experiment mixture of 80% helium and 20% air (mole fractions) is taken as the sample dual gas composition. A typical initial flow meter setting of 4.0 L P M is also used in these sample calculations. 155 F . l . Flow Rate Correction 156 F . l . l Direct Correction The flow rate of an actual gas (helium/air mixture in this case) is related to that of the reference/calibration gas by [22] . Qmix — ~, p7~\ * Qref (E-1) [P^pjmix where the thermal properties are taken at standard conditions. The actual gas density and specific heat of the mixture is based on the component mass fractions in the total flow. rhHe , rhair Pmix = ~. X pHe + — _ X Pair (F.2) rhT = mHe + rhair (F.3) The expression for the mixture Cp is similar to Equation F.2. The mass fractions corre-sponding to 80% and 20% mole fractions of helium and air are 0.8(4.003 g/mol) 0.8(4.003 g/mol) + 0.2(28.97 g/mol) ~ ( } 0.2(28.97 g/mol) M , , v »i > - .64.4% air (F.5) 0.8(4.003 g/mol) + 0.2(28.97 g/mol) Therefore, the respective mixture properties are calculated to be Pmix = 0.356(0.1615 kg/m3) + 0.644(1.169 kg/m3) - 0.8.10 kg/m3 (F.6) {Cp)mix = 0.356(5.193 kJ/kgK) + 0.644(1.004 kJ/kgK) = 2.495 kJ/kgK (F.7) Finally, the corrected actual flow rate of the dual gas mixture is (1.169 kg/m3)(l.004kJ/kgK) , = (0.8!0 4/m^)(2.495 J,kgK) ^ L P M ) ~ 2 3 2 3 L P M <R8» F . l . Flow Rate Correction 157 F.l.2 Manufacturer-Suggested Correction Alternatively, the manufacturer suggests a slightly different correction procedure that involves more complex C p and molecular structure expressions for the mixture [23]. The main equation relating the actual and reference gases is ^ = ( ^ ) { § r x ^ (F.9) • " " r e / {P^p/mix The mixture density is found by the same procedure as before, while the mixture C p is determined using a density-weighted calculation. ( C , U = ^ x ( C p ) f l . + ^ x ( C P W (F.10) The mixture molecular structure is calculated as Nmix = ^ x NHe + ^  x Nair ( F . l l ) TTIT TUT where N is 1.04 and 1.00 for monoatomic and diatomic gases, respectively. Therefore, using the previous calculated results for a 80/20 sample composition, the mixture C p and N are found to be = 1.302 U/kgK (F 12) Nmix = 0.356(1.04) + 0.644(1.00) = 1.014 (F.13) while the corrected mixture sample flow rate is .1.014,(1.169 kg/m3)(l.004 kJ/kgK) l A n T ^ l X A r r „ w / T , A, Q ™ " (Tmr )(0.81o4/m3)(1.302TO) * ( 4 ' ° L P M ) = 4 5 1 5 ( F 1 4 ) The significant difference between these two correction results, as well as their discrep-ancies to the measured volume flow rate (discussed in the next section) give motivation to experimentally calibrate the volume flow rate. This external audit is also important F.2. Experimental Calibration 158 because dissimilar thermal properties (between the sample dual gas mixture and the original calibration gas) often results in inaccurate theoretical corrections [23]. F.2 Experimental Calibration The purpose of the manual calibration is to obtain a convenient and accurate method to calculate actual sample gas flow rates from each shock tube experiment. The true volume flow rate depends on two main variables: helium fraction and flow meter setting. Although the range of the flow meter is 0-10 L P M , the internal diaphragm pump is only capable of (and thus limiting the maximum gas sample flow rate to) 6 L P M . Therefore the flow meter is initially set (using its controller) between 1.0 and 6.0 L P M , when sensing ambient air. Due to documented inaccuracies [10] over certain ranges, the best compromise is around 4 L P M to prevent automatic shutoffs when the sensed flow deviates from the initial setting by more than 10%. Helium fractions vary from 85% to 95% by volume, based on typical experimental conditions. The calibration apparatus setup is shown in Figure F . l , with a bubble-type volume flow meter as the reference flow device. The helium and air fractions (of the calibration gas) are controlled by the partial pressure method in the shock tube, before dumping the contents into the sample bag to create proper mixing. The sample flow must be connected to the aethalometer aerosol inlet (instead of outlet) due to possible small leakages through the quartz tape compression spot. bubble flow meter ambient air F igu re F . l : Flow Meter Calibration Apparatus The calibration results arc summarized in Table F . l , by averaging over several trials. It can be seen from Figure F.2 (using representative values) that the actual flow rates exhibit F.2. Experimental Calibration 159 very good linear correlations within the variable ranges of interest. The repeatability of the flow rates (with ambient air) over several days is within 3%. A Matlab program is used to determine the volume flow rate from a given helium fraction and flow meter setting, using a two-dimensional cubic spline interpolation as shown in Figure F.3. Finally, the raw calibration data is included in Table F.2 for completeness. Tab le F . l : Calibrated Flow Rates (LPM) Helium Fraction (vol. %) 100 90 80 0 Flow 2.0 2.92 2.69 2.49 1.84 Meter 3.0 3.96 3.82 3.54 2.73 Setting 4.0 4.93 4.68 4.53 3.68 (LPM) 5.0 5.65 5.44 5.28 4.55 F.2. Experimental Calibration 160 Tab le F .2 : Calibration Data Setting Volume Range Time Flow Rate Setting Volume Range Time Flow Rate (LPM) (mL) (s) (LPM) (LPM) (mL) (s) (LPM} Ambient Air 2.0 100-400 9.68 1.86 2.0 100-800 22.16 1.90 2.0 100-400 9.66 1.86 2.0 100-800 22.03 1.91 2.0 100-400 9.72. 1.85 2.0 100-800 22.12 1.90 2.0 100-400 9.69 1.86 2.0 100-800 22.25 1.89 2.0 100-400 9.63 1.87 2.0 100-800 22.06 1.90 3.0 100-400 6.65 2.71 3.0 100-800 15.19 2.76 3.0 100-400 6.59 2.73 3.0 100-800 15.07 2.79 3.0 100-400 6.62 2.72 3.0 100-800 15.12 2.78 3.0 100-400 6.66 2.70 3.0 100-800 15.09 2.78 3.0 100-400 6.53 2.76 3.0 100-800 15.09 2.78 4.0 100-400 5.00 3.60 4.0 100-800 11,31 3.71 4.0 100-400 4.91 3.67 4.0 100-800 11.40 3.68 4.0 100-400 5.03 3.58 4.0 100-800 11.41'" 3.68 4.0 100-400 4.97 3.62 4.0 100-800 11.31 3.71 4.0 100-400 4.94 3.64 4.0 100-800 11.32 3.71 5.0 100-400 4.06 4.43 5.0 100-800 9.18 4.58 5.0 100-400 4.03 4.47 5.0 100-800 9.19 4.57 .5.0 100-400 4.16 4.33 5.0 100-800 9.16 4.59 5.0 100-400 4.16 4.33 5.0 100-800 9.12 4.61 5.0 100-400 4.10 4.39 5.0 100-800 9.06 4.64 100% Helium 2.0 100-400 6.22 2.89 2.0 100-800 14.28 2.94 2.0 100-400 6.31 2.85 2.0 100-800 14.18 2.96 2.0 100-400 6.22 2.89 2.0 100-800 14.15 2.97 2.0 100-400 6.25 2.88 2.0 100-800 14.19 2.96 2.0 100-400 6.28 2.87 2.0 100-800 14.09 2.98 3.0 100-400 4.59 3.92 3.0 100-800 10.43 4.03 3.0 100-400 4.63 3.89 3.0 100-800 10.47 4.01 3.0 100-400 4.63 3.89 3.0 100-800 10.44 4.02 3.0 100-400 4.62 3.90 3.0 100-800 10.47 4.01 3.0 100-400 4.59 3.92 3.0 100-800 10.53 3.99 4.0 100-400 3.59 5.01 4.0 100-800 8.38 5.01 4.0 100-400 3.72 4.84 4.0 100-800 8.40 5.00 4.0 100-400 3.72 4.84 4.0 100-800 8.32 5.05 4.0 100-400 3.72 4.84 4.0 100-800 8.41 4.99 4.0 100-400 3.75 4.80 4.0 100-800 8.50 4.94 5.0 100-400 3.19 5.64 5.0 100-800 7.37 5.70 5.0 100-400 3.19 5.64 5.0 100-800 7.38 5.69 5.0 100-400 3.22 5.59 5.0 100-800 7.28 5.77 5.0 100-400 3.22 5.59 5.0 100-800 7.40 5.68 5.0 100-400 3.22 5.59 5.0 100-800 7.43 5.65 F.2. Experimental'Calibration 161 Setting (LPM) Volume Range (mL) Time Flow Rate Setting Volume Range Time Flow Rate| (s) (LPM) Ambient Air 2.0 100-400 9.91 1.82 2.0 100-800 2.0 100-400 9.94 1.81 2.0 100-800 2.0 100-400 9.85 1.83 2.0 100-800 2.0 100-400 9.94 1.81 2.0 100-800 2.0 100-400 10.00 1.80 2.0 100-800 3.0 100-400 6.62 2.72 3.0 100-800 3.0 100-400 6.65 2.71 3.0 100-800 3.0 100-400 6.71 2.68 3.0 100-800 3.0 100-400 6.59 2.73 3.0 100-800 3.0 100-400 6.60 2.73 3.0 100-800 4.0 100-400 4.94 3.64 4.0 100-800 4.0 100-400 4.94 3.64 4.0 100-800 4.0 100-400 s 4.91 3.67 4.0 100-800 4.0 100-400 4.90 3.67 4.0 100-800 4.0 100-400 4.87 3.70 4.0 100-800 22.75 22.72 22.78 22.66 22.60 15.09 15.16 15.13 15.10 15.13 11.31 11.22 11.25 11.19 11.19 1.85 1.85 1.84 1.85 1.86 2.78 2.77 2.78 2.78 2.78 3.71 3.74 3.73 3.75 3.75 5.0 100-400 3.94 4.57 5.0 100-800 9.10 4.62 5.0 100-400 3.97 4.53 5.0 100-800 9.10 ; 4.62 5.0 100-400 3.90 4.62 5.0 100-800 9.10 4.62 5.0 100-400 3.85 4.68 5.0 100-800 9.07 4.63 5.0 100-400 4.06 4.43 5.0 100-800 9.06 4.64 90% Helium 2.0 100-400 6.81 2.64 2.0 100-800 15.44 2.72 2.0 100-400 6.75 2.67 2.0 100-800 15.41 2.73 2.0 100-400 6.71 2.68 2.0 100-800 15.41 2.73 2.0 100-400 6.81 2.64 2.0 100-800 15.50 2.71 2.0 100-400 6.78 2.65 2.0 - 100-800 15.40 2.73 3.0 100-400 4.68 3.85 3.0 100-800 10.87 3.86 3.0 100-400 4.78 3.77 3.0 100-800 10.94 3.84 3.0 100-400 4.71 3.82 3.0 100-800 10.88 3.86 3.0 100-400 4.78 3.77 3.0 100-800 10.94 3.84 3.0 100-400 . 4.85 3.71 3.0 100-800 10.87 3.86 4.0 100-400 3.85 4.68 4.0 100-800 8.97 4.68 4.0 100-400 3.81 4.72 4.0 100-800 9.03 4.65 4.0 100-400 3.85 4.68 4.0 100-800 8.88 4.73 4.0 100-400 3.90 4.62 4.0 100-800 8.91 4.71 4.0 100-400 3.88 4.64 4.0 100-800 8.87 4.74 5.0 100-400 3.32 5.42 5.0 100-800 7.66 5.48 5.0 100-400 3.35 5.37 5.0 100-800 7.68 5.47 5.0 100-400 3.32 5.42 5.0 100-800 7.66 5.48 5.0 100-400 3.34 5.39 5.0 100-800 7.69 5.46 5.0 100-400 3.28 5.49 5.0 100-800 7.75 5.42 F .2 . Experimental Calibration i 162 1 Setting (LPM) Volume Range (mL) Time (s) Flow Rate (LPM) Setting (LPM) Volume Range (mL) Time (s) Flow Rate (LPM) Ambient Air 2.0 100-400 9.97 1.81 2.0 100-800 22.94 1.83 2.0 100-400 10.00 1.80 2.0 100-800 22.94 1.83 2.0 100-400 10.06 1.79 2.0 100-800 22.90 1.83 2.0 100-400 9.93 1.81 2.0 100-800 22.91 1.83 2.0 100-400 10.00 1.80 2.0 100-800 22.94 1.83 3.0 100-400 6.72 2.68 3.0 100-800 15.40 2.73 I 3.0 100-400 6.68 2.69 3.0 100-800 15.47 2.71 3.0 100-400 6.71 2.68 3.0 100-800 15.34 2.74 3.0 100-400 6.75 2.67 3.0 100-800 15.43 2.72 3.0 100-400 6.72 2.68 3.0 100-800 15.37 2.73 4.0 100-400 4.94 3.64 4.0 100-800 11.35 3.70 4.0 100-400 4.97 3.62 4.0 100-800 11.38 3.69 4.0 100-400 4.96 3.63 4.0 100-800 11.31 3.71 4.0 100-400 4.94 3.64 4.0 100-800 11.34 3.70 4.0 100-400 4.97 3.62 4.0 100-800 11.28 3.72 5.0 100-400 3.97 4.53 5.0 100-800 9.12 4.61 5.0 100-400 3.97 4.53 5.0 100-800 9.13 4.60 5.0 100-400 4.00 4.50 5.0 100-800 9.19 4.57 5.0 100-400 3.97 4.53 5.0 100-800 9.15 4.59 5.0 100-400 4.00 4.50 5.0 100-800 9.13 4.60 80% Helium 2.0 100-400 7.29 2.47 2.0 100-800 16.62 2.53 2.0 100-400 7.34 2.45 2.0 100-800 16.59 2.53 2.0 100-400 7.34 2.45 2.0 100-800 16.68 2.52 2.0 100-400 7.31 2.46 2.0 100-800 16.68 2.52 2.0 100-400 7.31 2.46 2.0 100-800 16.69 2.52 3.0 100-400 5.16 3.49 3.0 100-800 11.72 3.58 3.0 100-400 5.16 3.49 3.0 100-800 11.69 3.59 3.0 100-400 5.15 3.50 3.0 100-800 11.72 3.58 3.0 100-400 5.13 3.51 3.0 100-800 11.75 3.57 3.0 100-400 5.15 3.50 3.0 100-800 11.72 3.58 4.0 100-400 4.00 4.50 4.0 100-800 9.13 4.60 4.0 100-400 4.09 4.40 4.0 100-800 9.13 4.60 4.0 100-400 4.04 4.46 4.0 100-800 9.22 4.56 4.0 100-400 4.00 4.50 4.0 100-800 9.16 4.59 4.0 100-400 3.97 4.53 4.0 100-800 9.16 4.59 5.0 100-400 3.50 5.14 5.0 100-800 7.75 5.42 5.0 100-400 3.50 5.14 5.0 100-800 7.72 5.44 5.0 100-400 3.50 5.14 5.0 100-800 7.78 .5.40 5.0 100-400 3.50 5.14 5.0 100-800 7.84 5.36 5.0 100-400 3.47 5.19 5.0 100-800 7.78 5.40 F.2. Experimental Calibration 163 Flow Rate Correlation for 80% Helium 6.00 i < 2.00 1.00 0.00 -I , — , , , 1 1.0 2.0 3.0 4.0 5.0 6 0 Setting (LPM) Flow Rate Correlation at 5.0 LPM Setting 5.70 T : — 5.20 -1 , , n 1 80 85 90 95 100 Helium % F igu re F .2 : Linear Flow Rate Correlations F.2. Experimental Calibration 164 F igu re F .3 : 2-D Surface Interpolation Appendix G Premixed Experiment Data G.l Procedure Figures G . l and G.2 illustrate the incident shock velocity calculation from the dynamic pressure transducer signals. G.2 Results Tables G.1-G.9 summarize the experimental conditions and Aethalometer measurements for all premixed series. The full Aethalometer data for each experiment can be found by cross-referencing these tables with the appropriate raw data files, based on the date and time. Tab le G . l : Premixed Methane Series Experimental Test Matrix Temperature (K) 1050 1100 1150 1200 1250 1300 1350 0.2 1 0.3 * 0.5 S 0.7 u 1 .£ 1.3 fi 1 7 2 XXX X X X X X X X X X X X X X Note: each X represents one experiment; all experiments at 30 bar pressure 165 G.3. Error Analysis 166 1 0.5 0 -0.5 ( 0.5 ra o > a5-0.5 g 0 1 1 0.5 i o -0.5 10 5 0 -5 4 —r-10 12 14 10 12 14 6 8 Time (ms) 10 12 14 16 16 I i i i 1 1 I . . . . . . . . — • — : •— — . . . 1 i — . 1 i 1 i i 16 16 F igu re G . l : Dynamic Pressure Transducer Signals G.3 Error Analysis Combustion conditions are affected by experimental errors in driven gas pressures and incident shock velocity. Errors in fuel masses and E Q R are also influenced by measure-ment errors in driven gas pressures. On the other hand, errors in the Aethalometer B C mass results are introduced by experimental uncertainties in sample volume and flow rate, collection tape spot area, optical specific attenuation factor, and data modification algorithms. G.3. Error Analysis 167 y= 1.0351x- 0.2743 R 2=1 4.5 F igu re G.2: Incident Shock Velocity Calculation G.3.1 Driven Gas Composition The driven gas and fuel composition was calculated by successfully adding individual gas species into the evacuted driven section, while measuring their partial pressures using the Auto Tran 600D-117 vacuum sensor and Circuit-Test DMR-3600 multimeter. The vacuum sensor was calibrated using a zero and span calibration. The error in the output voltage (multimeter resolution) is ±0.01 V at vacuum and ±0.001 V at atmosphere. A n additional error of ±0.001 bar at atmospheric pressure stems from the Oakton Aneroid barometer uncertainty. Voltage errors of ±0.001 V in the partial pressure measurements translate into ±0.002 bar (±0.029 psi). The order in which the driven section was filled also intro-duced variable errors in pressures (and mole fractions) of individual gases. Therefore the various driven gas pressure errors (in bars) are summarized below [21]. ^ethane P'measured, 1 i 0.002 (G. l ) Pmethane ^measured,2 ^measured,! ± 0.004 (G.2) G.3. Error Analysis 168 Tab le G . 2 : Premixed Methane/Ethane Series Experimental Test Matrix Temperature (K) 1000 1100 1200 .2 0.5 13 1 cn ' o) 2 1 3 ro 4 > • . ' § • 5 LU 6 X X X X X X xooo X X + X X X X Note: X, 0, and + represent 40, 16, and 30 bar experiments, respectively Tab le G . 3 : Premixed Blank Series Experimental Test Matr ix Temperature (K) 1100 1200 -o 17 <D 5? 30 (/) <u * 40 X X xxxxx xxxxx X X X X X Note: each X represents one experiment Pair fmea3ured,3 Pmeasured,2 ± 0.004 Ptotal ~ Pmeasuredfi ±0.002 The errors in the individual mole fractions of the driven gases are therefore: Pethane Pmeasured,\ ± 0.002 Vethane Ptotal Pmeasured,3 ±0.002 _ ^methane _ Pmeasured,2 ~~ Pmeasured.l.^ 0.004 ymethane ;— D ~~ ~ — total *measured}3 ±0.002 _ Pair _ Pmeasured,3 ~ Pmeasured,2 ± 0.004 Vair — ~ j • •'total ^ measured,3 ± 0.002 (G.3) (G.4) (G.5) (G.6) (G.7) G.3. Error Analysis 169 Tab le G .4 : Premixed.Methane Series Conditions Methane Series Number Name Methane Equivalence Pressure Temperature Mass (g) Ratio (bar) (K) M1 pnopmOl n/a n/a 29.7 1194 M2 p30s01 0.433 1.02 27.7 1129 M3 p30s02 0.302 0.75 27.3 1186 M4 p30s03 0.212 0.53 31.3 1311 M5 p30s04 0.583 1.40 27.7 1100 , M 6 p30s05 0.125 0.32 28.7 1308 M7 p30s06 0.773 1.81 29.2 1074 M8 p30s07 0.915 2.10 29.6 1045 M9 p30s08 0.082 0.21 24.3 1214 M10 p30s09 0.082 0.21 22.6 1173 M11 p.30s10 0.082 0.21 22.5 1169 M12 p30s11 0.434 1.05 22.0 1026 M13 p30s12 0.438 1.06 23.9 1067 M14 p30s13 0.421 1.00 23.7 1060 M15 p30s14 0.422 1.00 23.3 1050 M16 p30s15 0.422 1.00 24.3 1073 Note: n/a indicates insufficient information was recorded for the calculation The errors in the equivalence ratios (for the respective series) are given by: 2(j^ methane) PmeasuredA ± 0.002 Methane EQR 0.21(t/ajr) Pmeasured,2 ~ Pmeasured,! ±0.004 (G.8) Methane/Ethane EQR = ^(Vmethane) + ^-^{y ethane) 0.21(y a i r) ^measured,?. ± 0.006 Pmeasuredfi Pmeasured,2 i 0.004 (G.9) For the average experiment in the methane series, the above errors resulted in methane and air pressure errors of 0.007 and 0.015 psi, respectively. These translated into er-rors of 0.0039 grams (1.84%) in the methane mass and 0.011 in the E Q R . For. the methane/ethane series, the average errors in methane, ethane, and air pressures are 0.007, 0.007, and 0.015 psi, respectively. This results in average mole fraction errors of 0.52% for both fuels. The corresponding methane and ethane fuel mass errors are 0.0039 and 0.0073 grams, respectively. The total fuel error is 0.0112 grams (1.11%). The resultant error in E Q R is calculated to be 0.023. G.3. Error Analysis 170 G.3.2 Experimental Temperature/Pressure Due to the ideal convergence of the linear transducer signal curve fit, only errors in relative transducer distances and signal sampling rates will be analyzed. Estimated uncertainties in the measured distances (±0.2%) and the radius of the transducer (0.002 m) contribute to the overall distance error, while the length of each sampling period (with sampling frequencies of 125 kHz per channel) is used as the uncertainty in the sampling rate. Ed = 0.002(dmea3ured) + 0.002 (G.10) £ ' = i 2 4 ( i = 8 x 1 0 " 6 ( G - n > The combined error in the incident shock velocity (V) is therefore dV dV Etotal = ^ ( — y { E d y + {-^y(Ety (G.12) Etotai = yj(^)2(0.002(dmeasured) + 0.002)2 + ( ^ ) 2 ( 8 x l O " 6 ) 2 (G.13) Using the largest relative transducer distance of 3.935 m and the calculated velocities, the uncertainty in the incident shock velocity ranges from 6.03-7.51 m/s for the methane series, 5.22-7.57 m/s for the methane/ethane series, and 4.68-6.21 m/s for the blank tests. Finally, using the error ranges of the driven gas pressures and shock velocities, the cumu-lative uncertainties in the experimental temperatures and pressures fall between 11-16 K and 0.5-0.7 bar for the methane series, 8-15 K and 0.4-0.8 bar for the methane/ethane series, and 7-11 K and 0.3-0.9 bar for the blank tests. G.3.3 Black Carbon Mass The most representative mass increment and the total sampling time required are used to determine the total B C mass. Measurement errors in the total sample volume and flow rate result in the total sampling time error. The sample volume error stems from the Eclipse high-pressure sensor's voltage output uncertainty of ±0.001 V (translated into 0.85 psi), while the volume flow rate error is estimated to be 2%. Although the particle collection spot area in the Aethalometer was stated as 0.5 cm 2 , its diameter is G.3. Error Analysis 171 manually measured to be 7.96 mm with a resolution uncertainty of 0.04 mm. Therefore these resultant area ranges will be used. Due to the known variablities in the specific attenuation factor [35], a conservative uncertainty in a of ±1.0 is estimated. v ^driver ~ ^measured ± 0.85 (G.14) Qsample ~ Qmeasured ± 0 '02(Q m e a s u r e c ( ) (G.15) Dtapespot = 7.96 mm ± 0.04 mm (G.16) a = a 'given ± 1-0 (G.17) Therefore, the combined total error ranges in the black carbon masses are found to be 9.4-10.1% for the methane series, 9.3-10.6% for the methane/ethane series, and 9.3-9.6% for the blank tests. Coupled with the fuel mass errors described previously, the normalized B C mass errors are 9.8-14.3% for the methane series and 10.0-12.3% for the methane/ethane series. Number Name Methane/Ethane Series Methane Ethane Methane Equivalence Mass(g) Mass (g) Fraction (mol %) Ratio Pressure (bar) Temperature! ME1 ME2 ME3 ME4 ME5 ME6 ME7 ME8 ME9 ME10 ME11 ME12 ME13 ME14 ME15 ME16 ME17 pnopm02 pnopm03 pnopm04 pnopm05 pnopm06 pnopmOT pnopm08 mepmOl mepm02 mepm03 mepm04 mepm05 mepm06 mepm14 mepm15 mepm16 mepm17 n/a n/a n/a n/a 0.472 0.565 0.698 0.565 0.902 1.157 1.326 1.550 0.577 0.318 0.576 0.897 1.172 n/a n/a n/a n/a 0.080 0.102 0.132 0.118 0.221 0.243 0.294 0.352 0.118 0.066 0.110 0.211 0.239 n/a n/a n/a n/a 91.7% 91.2% 90.9% 90.0% 88.5% 89.9% 89.4% 89.2% 90.2% 90.0% 90.7% 88.8% 90.2% n/a n/a n/a n/a 1.009 1.008 1.010 1.034 2.099 3.163 4.270 6.723 1.053 0.513 1.037 2.085 3.238 26.0 18.0 16.9 17.1 38.2 41.3 41.7 34.3 35.5 34.6 39.0 34.0 38.4 37.9 36.0 36.8 40.6 1125 1270 1182 1239 1174 1117 1018 1024 1044 1034 1089 1031 1079 1072 1047 1061 1107 Note: n/a indicates insufficient information was recorded for the calculation G.3. Error Analysis 173 Tab le G . 6 : Premixed Blank Series Conditions tn c 0) E E o O 10 <D CO IW C TO > m cu 3 (0 o3 * Q . E CU cu E TO Z 0) XI E 3 Z tt) > T3 C o c 0 O co LU T -LU o 5 P <D TO TO TO TO T3 X J O O O O 0) d) o o c c •g "g T3 •<- 00 > c c TO a> u x: 03 3 to to to in in in E E ' E E E E 0 0 ) 0 ) 0 ) 0 ) 0 ) TO TO TO TO TO TO i _ L _ u . L . k_ i _ x: x: x: x: x: x: O . Q . Q . O . Q . Q _ c TO TO TO TO TO TO o T5 T3 t ) T J b 13 C TO X LU a> TO a> TO 0) a) TO TO r - t (O CO O B (<) n m io o to N s O T - O T - o o o a> co m a> o co •<--*r h-' o CD oo od co Ti- ro CO 5 2 ? CO o a> o <o co i n m csi i r i i r i i r i r~-' T - C N t N C M C M C N C N C N ° N O O C B O T - ( < ) T - ( M ( < ) T f i n ( O N 0 3 0 1 0 O O O T - T - T - O O O O O O O O O T -£ S= p C C C X ) X I X 1 X I X 1 X 1 X I X I X I X ) o o S n n o w < / ) < ' > < / > < ' > i o t o < o i o t o c o c o c o c o c o c o c o c o c o c o E E E E E E Q . Q . Q . Q . Q . Q . a. a a a 3 s "D cu TO in cu </) cu ^ "S 3 8 % cu t: •- o to -= TO .55 ^ c C 0) O 3 ~ CT TO CU E w •- cu c E .9> o ,9 t £ fi — TO cS J2 l § xs E | 8-c a) o z CN Number Name Aethalometer Date Methane Series File (m/d/y) M1 pnopmOl BC160903 9/16/03 M2 p30s01 BC291003 10/29/03 M3 p30s02 BC301003 10/30/03 M4 p30s03 BC301003 10/30/03 M5 p30s04 BC311003 10/31/03 M6 p30s05 BC051103 11/5/03 M7 p30s06 BC061103 11/6/03 M8 p30s07 BC071103 11/7/03 M9 p30s08 BC171103 11/17/03 M10 p30s09 BC181103 11/18/03 M11 p30s10 BC191103 11/19/03 M12 p30s11 BC201103 11/20/03 M13 p30s12 BC211103 11/21/03 M14 p30s13 BC251103 11/25/03 M15 p30s14 BC261103 11/26/03 M16 p30s15 BC271103 11/27/03 Sample Duration Initial ATN Total Volume Flow Rate Sample Time BC Mass Increment Total BC Mass Normalized BC Mass (L) (LPM) (min) (ng/min) (ng) (uq/q) 17:44-18:51 4.62 351.6 4.76 73.8 0.607 45 n/a 16:47-17:35 3.345 343.7 4.76 72.2 393.119 28370 65.465 13:11-14:12 3.921 346.1 4.78 72.4 414.613 30032 99.331 17:04-17:41 6.978 353.1 4.85 72.8 286.039 20826 98.218 13:10-14:35 7.015 348.7 4.82 72.3 195.663 14149 24.275 13:27-14:10 13.495 351.5 4.86 72.3 114.758 8293 66.204 14:38-15:21 "7.139 353.0 4.81 73.4 135.583 9951 12.865 15:07-15:39 1.925 349.2 4.79 72.8 179.712 13087 14.299 16:55-17:32 13.51 356.6 4.87 73.2 20.790 1522 18.551 13:26-14:07 12.392 357.7 4.87 73.4 33.082 2430 29.619 14:13-14:54 1.805 365.0 4.87 74.9 131.068 9820 119.777 13:44-14:26 3.412 362.6 4.77 76.0 88.603 6733 15.515 13:20-14:02 19.186 360.2 4.76 75.6 69.782 5277 12.055 15:58-16:38 18.404 352.1 4.77 73.8 54.502 4023 9.552 16:16-17:02 10.736 345.9 4.77 72.5 47.629 3455 8.186 14:38-15:20 0.226 341.1 4.77 71.5 42.297 3023 7.161 w cr ST O T I ra 3 x ra &-cr S ra cc ra —i ra" c n > ra cr o" B a ra i-( O o 0 Note: a = 16.6 m2/g; collection spot area = 0.5 cm 2 Number Name Aethalometer Date File (m/d/y) Methane/Ethane Series Sample Duration Initial Total Flow Sample BC Mass Total BC Normalized ATN Volume Rate Time Increment Mass BC Mass (D (LPM) (min) (ng/min) (ng) (ua/a^ 31.617 346.6 4.79 72.3 5.445 V '3 / 394 \ u y a; n/a 52.467 227.5 4.79' 47.5 4.414 210 n/a 15.347 227.5 4.79 47.5 1.882 89 n/a 26.615 220.9 4.62 47.8 2.079 99 n/a 10.56 451.2 4.79 94.2 5.357 504 0.913 119.786 452.3 4.78 94.7 6.556 621 0.929 1.002 417.0 4.69 88.8 10.647 946 1.140 9.106 469.3 4.72 99.3 139.563 13863 20.316 6.712 468.1 4.82 97.2 23.149 2250 2.005 29.819 468.5 4.85 96.5 22.735 2195 1.568 71.903 467.9 4.88 95.8 37.803 3623 2.237 18.023 461.2 4.87 94.6 1155.717 109360 57.487 16.696 476.3 4.77 99.9 1439.000 143763 206.790 36.881 473.8 4.73 100.2 132.607 13281 34.609 2.356 473.0 4.72 100.2 144.098 14438 21.049 1.802 471.7 4.81 98.0 249.109 24421 22.030 0.973 470.8 4.85 97.1 4404.331 427574 302.909 ME1 ME2 ME3 ME4 ME5 ME6 ME7 ME8 ME9 ME10 ME11 ME12 ME13 ME14 ME15 ME16 ME17 pnopm02 pnopm03 pnopm04 pnopm05 pnopm06 pnopm07 pnopm08 meprnOI mepm02 mepm03 mepm04 mepm05 mepm06 mepm14 mepm15 mepm16 mepm17 BC170903 BC170903 BC180903 BC180903 BC190903 BC190903 BC190903 BC031003 BC031003 BC061003 BC061003 BC071003 BC081003 BC181003 BC191003 BC191003 BC191003 9/17/03 9/17/03 9/18/03 9/18/03 9/19/03 9/19/03 9/19/03 10/3/03 10/3/03 10/6/03 10/6/03 10/7/03 10/8/03 10/18/03 10/19/03 10/19/03 10/19/03 12:12 15:48 12:29-14:42-11:34-14:06-17:04-14:14-17:16-14:32-17:09-14:14-13:16-16:03-11:41 13:29 15:16 •13:20 16:50 13:29 15:35 13:15 15:48 18:30 15:38 18:23 15:25 17:59 14:30 14:16 17:01 12:29 14:36 15:48 Note: a = 16.6 m2/g; collection spot area = 0.5 cm 2 Number Name Aethalometer Date Sample File (m/d/y) Duration Blank Initial ATN Series Total Volume Flow Rate (LPM) Sample Time (min) BC Mass Increment (ng/min) Total BC Mass (ng) Normalized | BC Mass (ng) B1 pnopmlO BC290903 9/29/03 16:46-17:35 10.77 251.7 B2 mepm07 BC091003 10/9/03 14:55-16:18 2.781 472.4 B3 mepm08 BC101003 10/10/03 14:14-14:41 14.263 463.1 B4 mepm09 BC141003 10/14/03 17:21-18:13 73.919 460.6 B5 mepmlO BC151003 10/15/03 16:13-16:45 10.332 256.6 B6 mepm11 BC161003 10/16/03 16:40-17:30 80.009 468.1 B7 mepm13 BC181003 10/18/03 14:11-15:10 9.968 468.1 B8 p30sb01 BC231003 10/23/03 12:32-13:36 86.682 441.1 B9 p30sb02 BC291003 10/29/03 13:26-14:06 9.152 345.2 B10 p30sb03 BC131103 11/13/03 12:43-13:30 20.969 343.5 B11 p30sb04 BC141103 11/14/03 12:37-13:26 18.759 345.3 B12 p30sb05 BC191103 11/19/03 17:02-17:52 124.69 346.3 B13 p30sb06 BC041203 12/4/03 14:17-15:02 9.159 344.9 B14 p30sb07 BC041203 12/4/03 17:04-17:40 62.146 344.9 B15 p30sb08 BC121203 12/12/03 16:49-17:30 6.301 346.2 B16* p30sb09 BC171203 12/17/03 16:55-17:35 5.591 338.6 B17 p30sb10 BC181203 12/18/03 12:21-12:59 11.626 341.2 4.62 4.86 4:70 4.70 4.63 4.70 4.70 4.68 4.72 4.72 4.72 4.77 4.72 4.72 4.72 4.71 4.72 54.4 97.2 98.5 98.1 55.5 99.5 99.5 94.3 73.2 72.8 73.2 72.5 73.1 73.1 73.4 71.8 72.3 132.712 333.781 207.253 60.359 35.536 26.358 15.658 142.730 121.272 65.881 23.519 32.538 41.589 56.801 27.766 78.918 50.028 7222 32443 20418 5919 1971 2623 1558 13459 8879 4796 1722 2361 3039 4155 2037 5668 3617 10044 24035 15431 4498 2688 1961 1165 10680 9002 4886 1746 2386 3084 4216 2059 5860 3711 Note: a = 16.6 m2/g; collection spot area = 0.5 cm2 Appendix H Non-Premixed Experiment Data H . l Procedure Figure H. l shows the initial J-43 injector (with 1.1 mm tip) mass flow characterization [20]. The averaged normalized mass flow rate is used in the data analysis. H.2 Results Tables H . l - H . l l summarize the experimental conditions and Aethalometer measurements for all non-premixed series. The full Aethalometer data for each experiment can be found by cross-referencing these tables with the appropriate raw data files, based on the date and time. H.3 Error Analysis The average driven gas (containing only air) pressure error in the non-premixed case is due to the vacuum sensor voltage output error, and is calculated using Equation H.l to be 8.647 psi. Using Equations G.10 to G.13, the uncertainty in the incident shock velocity ranges from 5.98-7.40 m/s for the methane series and 6.04-7.36 m/s for the methane/ethane series. Therefore, the overall experimental temperature and pressure uncertainties fall between 11-15 K and 0.57-0.64 bar for the methane series, and 11-14 K and 0.59-0.64 bar for the methane/ethane series. The overall B C mass errors, using Equations G.14 to G.17, range from 9.5-17.6% for the methane/DME series, 9.4-9.5% for 177 H.3. Error Analysis 178 tn E f CO tn D-Measured Final Mass in Test Chamber Injection Pressure (psig) 834 1014 1234 0.5 1.790482666 2.046109308 2.338567234 1 3.800053003 4.384676542 5.407767735 1.5 5.700079504 6.138547159 8.330885429 J-43 Mass Flow Correlation Pulse Width (ms) F igu re H . l : J-43 Mass Flow Correlation the methane series, and 9.5-10.3% for the methane/ethane series. Since fuel mass errors will not be quantified, these also represent the normalized B C mass uncertainties. Pdriven — ^measured ± 0.002 (H.l) H.3. Error Analysis 179 Tab le H . l : Non-Premixed Methane/DME Series Experimental Test Matr ix 1100 1200 Temperature (K) 1300 1400 1500 1600 7 O XXX* O O O O CD CD k_ 10 X OO OO O O O O O 3 in m £ 20 X XX 0 Q- 25 X 0 0 0 0 0 O Note: 1) X represents injection experiment; 0 represents blank experiment 2) Injection Duration = 1 ms (* denotes 0.5ms injection) 3) Injection Pressure = 50 bar Tab le H.2 : Non-Premixed Methane Series Experimental Test Matr ix Temperature (K) 1100 . 1150 1200 1250 1300 1350 1400 = 30 bar xxxxx xxxx XX XXXXX xxxxx XXX XXX Pressure 0 0 0 OOOOO OOOOO OOOOO Note: 1) X represents injection experiment; 0 represents blank experiment 2) Injection Duration = 1.5 ms 3) Injection Pressure = 75 bar H.3. Error Analysis 180 Tab le H .3 : Non-Premixed Methane/Ethane Series Experimental Test Matr ix Temperature (K) 1100 1150 1200 1250 1300 1350 1400 30 bar XX XX XX XX XX X XX XX XX XX Pressure = OO O oo oo oo 0 Note: 1) X represents injection experiment; 0 represents blank experiment 2) Injection Duration = 2.6 ms 3) Injection Pressure = 120 bar Number Name Fuel Mass Pressure Temperatu (mg) (bar) (K) NMD1 NMD2 dmeinj03 dmeinj04 3.098 3.098 21.2 21.2 1177 1176 NMD3 dmeinj05 3.098 17.6 1074 NMD4B dmeinj07 0 21.4 1184 NMD5B dmeinj08 0 24.0 1489 NMD6B dmeinj09 0 22.9 1226 NMD7B dmeinjlu 0 22.5 1213 NM08B dmeinjH 0 22.1 1202 NMD9B dmeinj12 0 22.7 1219 NMD10B dmeinj13 0 22.7 1219 NMD11B dmeinj14 0 10.7 1455 NMD12B dmeinj15 0 4.4 1255 NMD13 dmeinj16 3.098 7.5 1671 NMD14B dmeinj17 0 7.4 1664 NMD15 dmeinj18 1.549 7.7 1701 NMD16 dmeinj19 3.098 7.5 1635 NMD17B dmeinj20 0 7.6 1675 NMD18B dmeinj21 0 7.7 1692 NMD19 dmeinj23 3.098 24.4 943 NMD20 dmeinj29 3.098 10.1 1308 NMD21B dmeinj30 0 8.6 1202 NMD22B dmeinj31 0 10.4 1327 NMD23B dmeinj32 0 10.3 13.19 NMD24B dmeinj35 0 10.5 1648 NMD25B dmeinj36 0 10:5 1648 NMD26B dmeinj37 0 10.7 1665 NMD27B dmeinj41 0 10.2 1413 NMD28B dmeinj42 0 10.8 . 1655 NMD29B dmeinj43 0 11.1 1676 N o n - P r e m i x e d M e t h a n e / D M E S e r i e s Apparatus Comments Cleaning Diaphragm Material start of bag sampling removed rubber gaskets dry cloth wet cloth inert driven gas (N2) inert driven gas (N2) filtered driven gas pressurized air (plastic tubing) pressurized air (plastic tubing) brush, Simple Green, water brush, Simple Green soak, water brush brush brush brush brush pressurized air (multiple jet plastic tubing) H 2 burnout, positive pressure cleaning filtered pressurized air minimize vacuum pump oil condensation grooved SS grooved SS grooved SS grooved SS aluminum grooved SS aluminum grooved SS grooved SS grooved SS plastic film plastic film plastic film plastic film plastic film plastic film plastic film plastic film grooved SS SS shim SS shim SS shim SS shim SS shim SS shim SS shim SS shim SS shim SS shim Note: 1) suffix B indicates a blank experiment 2) target injection durations of 1 ms (0.5 3) approximate injection pressures of 50 ms for NMD15) will be used since the injection signal was not recorded bar will be used Non-Premixed Methane Series Number Name Fuel Mass Pressure Temperature Injection Apparatus Comments (mg) (bar) (K) Pressure (bar) NM1 NM2 p30nt01 p30nt04 7.666 8.224 30.6 30.0 1422 1300 75 75 10-minute bag settling before sampling NM3 p30nt05 8.480 30.6 1210 . 75 NM4 p30nt06 8.368 29.2 1133 75 NM5 p30nt07 7.624 29.3 1293 75 NM6 p30nt08 7.964 28.9 1152 75 cr NM7B p30nt09b 0 29.0 1178 n/a ST NM8B p30nt10b 0 28.5 1168 n/a NM9B p30nt11b 0 28.5 1168 n/a Or NM10 p30nt12 7.773 28.3 1262 75 "Z NM11 p30nt13 7.996 28.5 1265 75 o a NM12 p30nt15 7.629 29.8 1293 74 i—t NM13 p30nt16 7.519 29.7 1293 74 a> a NM14 p30nt17 8.033 29.7 1293 74 x' NM15 p30nt18 7.352 30.3 1305 72 CD D -NM16 p30nt19 7.624 30.4 1307 75 NM17 p30nt20 8.331 30.1 1300 75 o c-t-NM18 p30nt21 7.959 29.1 1280 75 0 NM19 p30nt22 7.773 29.7 1293 75 e Ser NM20 p30nt23 7.885 30.1 1300 75 e Ser NM21B p30nt24b 0 29.9 1300 n/a CB' CO NM22B NM23B p30nt26b p30nt27b 0 0 30.2 30.0 1305 1298 n/a n/a start of 1/2" single-stage impactor Con NM24B p30nt28b 0 30.1 1300 n/a a. c-t-' NM25B p30nt29b 0 29.8 1293 n/a 0' NM26 p30nt30 7.662 29.7 1142 75 CO NM27 p30nt31 8.108 30.5 1158 75 NM28 p30nt32 7.662 30.4 1157 75 NM29 p30nt33 8.220 29.8 1147 75 NM30 p30nt34 8.034 29.6 1143 75 NM31 p30nt35 7.550 30.1 1153 75 NM32 p30nt36 7.550 28.2 1353 75 NM33 p30nt37 7.671 30.2 1402 75 Number Name Fuel Mass Pressure Temperature Injection (mg) (bar) (K) Pressure (psi) NM34 p30nt38 7.848 30.7 1216 75 NM35 p30nt39 7.997 30.2 1152 75 NM36B p30nt40b 0 29.7 1293 n/a NM37B p30nt41b 0 30.1 1300 n/a NM38B p30nt42b 0 29.8 1293 n/a NM39B p30nt43b 0 30.0 1300 n/a NM40B p30nt44b 0 30.3 1305 n/a NM41B p30nt45b 0 30.4 1307 n/a NM42B p30nt46b 0 30.3 1307 n/a NM43B p30nt47b 0 28.5 1263 n/a NM44B p30nt48b 0 30.6 1312 n/a NM45B p30nt49b 0 30.7 • 1312 n/a Apparatus Comments Note: 1) suffix B indicates a blank experiment 2) 0.030" Lexan diaphragms used for all experiments 3) n/a = not applicable start of 3-stage impactor non-cond. bag; back to 1/2" single-stage imp. Simple Green/water cleaning no impactor Number Name Fuel Mass Pressure Temperature Injection (mg) (bar) (K) Pressure (bar) NME1B p30nts04b 0 29.6 1288 n/a NME2B p30nts05b 0 29.8 1293 n/a NME3B p30nts13b 0 30.3 1307 n/a NME4B p30nts19b 0 30.0 1300 n/a NME5B p30nts22b 0 30.0 1300 n/a NME6B p30nts23b 0 29.5 1288 n/a NME7B p30nts26b 0 29.2 1133 n/a NME8B p30nts30b 0 29.9 1147 n/a NME9B p30nts35b 0 29.6 1142 n/a NME10 p3ntsb01 1.234 30.0 1300 120 NME11 p3ntsb03 1.227 29.5 1288 120 NME12 p3ntsb04 1.231 29.2 1281 120 NME13 p3ntsb05 1.231 30.0 1200 120 NME14 p3ntsb06 1.231 29.7 1194 120 NME15 p3ntsb07 1.231 30.3 1204 120 NME16 p3ntsb08 1.231 30.0 1200 120 NME17 p3ntsb09 1.231 29.8 1397 120 NME18 p3ntsb10 1.231 30.0 1395 120 NME19 p3ntsb11 1.231 29.8 1393 120 NME20 p3ntsb12 1.234 30.0 1400 120 NME21 p3ntsb13 1.231 30.1 1347 120 NME22 p3ntsb14 1.234 30.4 1359 120 NME23 p3ntsb15 1.234 29.6 1339 120 NME24 p3ntsb16 1.234 29.9 1346 120 NME25 p3ntsb17 1.231 30.1 1252 120 NME26 p3ntsb18 1.234 29.3 1233 120 NME27 p3ntsb19 1.231 30.0 1250 120 NME28 p3ntsb20 1.234 30.1 1250 120 Apparatus/Procedure Comments Note: 1) suffix B indicates a blank experiment 2) 0.030" Lexan diaphragms used for all experiments 3) * indicates settling duration in the tube prior to sampling 4) n/a = not applicable start of multi-jet impactor *60 minutes start of 1" multi-jet impactor *50 minutes *40 minutes *30 minutes *60 minutes *20 minutes *60 minutes *60 minutes *60 minutes *60 minutes *40 minutes *60 minutes *60 minutes *60 minutes *60 minutes *60 minutes *60 minutes. *60 minutes *60 minutes *60 minutes *60 minutes *60 minutes *40 minutes V cr ST •z 3 i T) 3 P ra c-t-tr 3 a> Ul CD >-! c o ' co o O 3 Q-O 3 co 9° >-! —I o > 3 co co' OO 4^ Number Name Aethalometer File N o n - P r e m i x e d M e t h a n e / D M E S e r i e s Date Sample Initial Total Flow Sample BC Mass Total BC Normalized (m/d/y) Duration ATN Volume Rate Time Increment Mass BC Mass (L) (LPM) (min) (ng/min) (ng) (uq/q) 7/10/03 15:36-16:17 1.893 239.7 4.70 51.0 3.411 ' 174 \"a*a/ 56.136 7/11/03 13:10-13:50 7.432 252.2 4.70 53.6 4.032 216 69.829 7/11/03 16:21-16:36 19.222 246.8 4.57 • 54.0 3.131 169 54.604 7/14/03 13:17-13:53 0.148 255.8 4.70 54.4 5.486 298 n/a 7/16/03 10:31-11:17 4.692 273.6 4.83 56.7 6.765 383 n/a 7/16/03 12:48-13:18 3.837 254.0 4.78 53.1 1.791 95 n/a 7/16/03 15:24-16:01 13.473 277.2 4.78 58.0 10.947 635 n/a 7/17/03 11:23-12:09 0.928 268.3 4.78 56.1 3.860 217 n/a 7/17/03 16:05-16:49 9.175 264.7 4.70 56.3 3.881 219 n/a 7/18/03 15:13-15:47 9.953 270.1 4.70 57.4 3.452 198 n/a 7/18/03 15:58-16:17 13.968 118.1 4.76 24.8 7.042 175 n/a 7/18/03 17:02-17:14 18.58 69.9 4.72 14.8 9.082 134 n/a 7/21/03 11:21-11:40 1.995 93.1 4.84 19.2 2.621 50 16.283 7/21/03 12:58-13:16 4.67 93.1 4.84 19.2 3.006 58 n/a 7/21/03 14:52-15:10 7.861 91.3 4.84 18.9 2.153 41 26.231 7/21/03 16:33-16:57 11.363 94.9 4.75 20.0 3.467 69 22.333 7/22/03 10:51-11:08 0.068 93.1 4.92 18.9 3.237 61 n/a .7/22/03 12:00-12:21 2.297 93.1 . 4.84 19.2 2.155 41 n/a 7/24/03 12:04-12:46 0.497 273.6 4.60 59.5 3.9793 237 76.478 7/25/03 11:15-11:36 0.854 114.5 4.78 23.9 13.7536 329 106.326 7/25/03 13:25-13:45 9.922 116.3 4.74 24.6 12.629 310 n/a 7/25/03 17:08-17:32 20.283 123.5 4.78 24.1 14.845 358 n/a 7/28/03 12:29-12:50 2.167 123.5 4.78 24.1 10.3347 249 n/a 7/30/03 11:17-11:42 1.088 123.5 4.95 23.3 5.6996 133 n/a 7/30/03 15:26-15:44 10.565 127.1 4.95 24.0 7.1378 171 n/a 7/30/03 16:36-16:55 16.18 ' 123.5 4.95 23.3 7.6321 178 n/a 7/31/03 16:04-16:26 35.596 112.8 4.91 23.0 11.7984 271 n/a 8/1/03 15:12-15:33 15.027 121.7 4.95 23.0 18.193 418 n/a 8/1/03 17:08-17:27 27.943 116.3 4.95 21.9 4.788 105 n/a NMD1 dmeinj03 BC100703 NMD2 dmeinj04 BC110703 NMD3 dmeinj05 BC110703 NMD4B dmeinj07 BC140703 NMD5B dmeinj08 BC160703 NMD6B dmeinj09 BC160703 NMD7B dmeinjIO BC160703 NMD8B dmeinjH BC170703 NMD9B dmeinj12 BC170703 NMD10B dmeinj13 BC 180703 NMD11B dmeinj14 BC180703 NMD12B dmeinj15 BC180703 NMD13 dmeinj16 BC210703 NMD14B dmeinj17 BC210703 NMD15 dmeinj18 BC210703 NMD16 dmeinj19 BC210703 NMD17B dmeinj20 BC220703 NMD18B dmeinj21 BC220703 NMD19 dmeinj23 BC240703 NMD20 dmeinj29 BC250703 NMD21B dmeinj30 BC250703 NMD22B dmeinj31 BC250703 NMD23B dmeinj32 BC280703 NMD24B dmeinj35 BC300703 NMD25B dmeinj36 BC300703 NMD26B dmeinj37 BC300703 NMD27B dmeinj41 BC310703 NMD28B dmeinj42 BC010803 NMD29B dmeinj43 BC010803 Note: 1) a = 16.6 m2/g; collection spot area = 0.5 cm2 2) n/a = not applicable Number Name Aethalometer Date Non-Premixed Methane Series Sample Initial File (m/d/y) Duration ATN NM1 . p30nt01 BC030304 3/3/04 17:03-17:20 45.371 NM2 p30nt04 BC050304 3/5/04 10:48-11:25 0.209 NM3 p30nt05 BC080304 3/8/04 11:41-12:23 12.758 NM4 p30nt06 BC080304 3/8/04 16:48-17:24 3.326 NM5 p30nt07 BC100304 3/10/04 11:36-12:07 8.934 NM6 p30nt08 BC100304 3/10/04 15:52-16:27 2.206 NM7B p30nt09b BC110304 3/11/04 13:26-13:56 7.672 NM8B p30nt10b BC110304 3/11/04 14:22-15:00 0.881 NM9B p30nt11b BC160304 3/16/04 11:13-11:51 6.822 NM10 p30nt12 BC160304 3/16/04 14:39-15:16 12.235 NM11 p30nt13 BC170304 3/17/04 10:15-10:51 9.937 NM12 p30nt15 BC180304 3/18/04 10:35-11:15 1.131 NM13 p30nt16 BC180304 3/18/04 12:05-12:43 0.687 NM14 p30nt17 BC190304 3/19/04 10:56-11:31 7.392 NM15 p30nt18 BC230304 3/23/04 11:38-12:01 11.124 NM16 p30nt19 BC230304 3/23/04 15:37-16:16 14.553 NM17 p30nt20 BC230304 3/23/04 -16:51-17:29 1.838 NM18 p30nt21 BC240304 3/24/04 11:10-11:46 2.811 NM19 p30nt22 BC240304 3/24/04 16:00-16:33 5.615 NM20 p30nt23 BC240304 3/24/04 16:52-17:31 0.139 NM21B p30nt24b BC250304 3/25/04 13:15-13:53 2.303 NM22B p30nt26b BC250304 3/25/04 16:22-16:58 10.916 NM23B p30nt27b BC290304 3/29/04 10:08-10:39 6.404 NM24B p30nt28b BC290304 3/29/04 10:58-11:22 15.11 NM25B p30nt29b" BC290304 3/29/04 11:39-12:00 0.27 NM26 p30nt30 BC310304 3/31/04 11:13-11:46 10.814 NM27 p30nt31 BC310304 3/31/04 16:38-17:09 3.413 NM28 p30nt32 BC310304 3/31/04 17:37-18:09 7.008 NM29 p30nt33 BC060404 4/6/04 16:45-17:00 10.37 NM30 p30nt34 BC070404 4/7/04 11:34-11:44 11.33 NM31 p30nt35 BC070404 4/7/04 16:26-16:55 30.146 NM32 j30nt36 BC080404 4/8/04 13:24-13:47 1.754 Total Flow Sample BC Mass Total BC Normalized Volume Rate Time Increment Mass BC Mass (L) (LPM) (min) (ng/min) (ng) (ua/a) 336.4 4.79 70.2 18.464 v a/ 1296 169.086 345.7 4.81 71.8 117.558 8441 1026.316 333.3 4.77 69.9 19.766 1381 162.887 332.9 4.71 70.7 13.251 937 111.948 338.0 4.81 70.2 17.247 1211 158.835 332.7 4.73 70.4 53.509 3766 472.928 348.8 4.63 75.4 13.021 981 n/a 348.0 4.71 73.8 14.219 1050 n/a 348.7 4.71 74.0 14.240 1053 n/a 348.9 4.76 73.2 12.405 908 116.868 350.3 4.76 73.6 16.566 1219 152.478 355.7 4.81 74.0 12.778 946 123.953 357.2 4.81 74.3 16.453 1223 162.599 350.9 4.81 73.0 15.658 1143 142.333 353.4 4.81 73.5 13.051 960 130.517 353.4 4.81 73.5 35.665 2622 343.916 353.4 4.81 73.5 28.806 2118 254.215 354.7 4.81 73.8 21.111 1558 195.754 354.0 4.81 73.6 17.524 1291 166.033 354.7 4.81 73.8 24.283 1792 227.292 354,7 4.72 75.1 27.576 2071 n/a 354.7 4.81 73.8 9.316 688 n/a 351.9 4.73 74.4 6.009 447 n/a 352.7 4.81 73.4 4.785 351 n/a 351.9 4.89 72.0 6.914 498 n/a 349.3 4.73 73.8 4.083 301 39.342 350.1 4.73 74.0 3.418 253 31.189 350.0 4.65 75.3 2.705 204 26.570 351.0 4.73 74.3 2.593 192 23.420 349.3 4.73 73.8 5.443 402 50.018 348.5 4.65 74.9 6.431 482 63.838 352.6 4.81 73.3 5.512 404 53.540 Number Name Aethalometer Date File (m/d/y) IN M O O NM34 p30nt37 p30nt38 BC080404 BC080404 4/8/04 4/8/04 NM35 p30nt39 BC080404 4/8/04 NM36B p30nt40b BC 120504 5/12/04 NM37B p30nt41b BC130504 5/13/04 NM38B p30nt42b BC140504 5/14/04 NM39B p30nt43b BC140504 5/14/04 NM40B p30nt44b BC170504 5/17/04 NM41B p30nt45b BC 180504 5/18/04 NM42B p30nt46b BC 180504 5/18/04 NM43B p30nt47b BC190504 5/19/04 NM44B p30nt48b BC200504 5/20/04 NM45B p30nt49b BC200504 5/20/04 Non-Prerrnxed Methane Sample Initial Duration ATN Series (Continued) Total Flow Volume Rate (L) (LPM), Sample Time (min' BC Mass Increment 'min) i ) (ng/Total BC Mass (ng) Normalized BC Mass o i -d CD X CD ra ra > ra cr SL o* B ra c-t-CD •-! o o o . 14:51-15:21 16:03-16:37 17:05-17:28 13:25-14:13 15:16-15:50 12:31-13:05 14:12-15:02 10:57-11:34 11:46-12:14 13:01-13:36 11:55-12:21 13:19-14:01 15:19-15:50 Note: 1) a = 16.6 m2/g; collection spot area 2) n/a = not applicable : 0.5 cm 2 6.3 0.459 5.275 3.919 11.086 5.653 5.496 7.225 3.806 1.623 5.366 28.776 1.881 353.4 351.1 349.7 352.7 352.0 350.9 350.9 351.3 352.0 351.3 351.4 350.9 350.9 4.85 72.8 5.946 433 56.454 4.77 73.7 4.813 355 45.181 4.73 74.0 7.021 519 64.951 4.80 73.5 9.896 727 n/a 4.80 73.3 22.510 1651 n/a 4.80 73.1 19.137 1399 n/a 4.80 73.1 48.875 3573 n/a 4.80 73.2 13.926 1019 n/a 4.80 73.3 10.320 757 n/a 4.80 73.2 22.156 1621 n/a 4.88 72.0 16.104 1159 n/a 4.80 73.1 22.287 1629 n/a 4.80 73.1 14.287 1044 n/a Number Name Aethalometer Date File; (m/d/y) Non-Premixed Methane/Ethane NME1B p30nts04b BC110604 NME2B p30nts05b BC110604 NME3B p30nts13b BC220604 NME4B p30nts19b BC240604 NME5B p30nts22b BC250604 NME6B p30nts23b BC250604 NME7B p30nts26b BC280604 NME8B p30nts30b BC290604 NME9B p30nts35b BC300604 NME10 p3ntsb01 BC180804 NME11 p3ntsb03 BC190804 NME12 p3ntsb04 BC190804 NME13 p3ntsb05 BC190804 NME14 p3ntsb06 BC230804 NME15 p3ntsb07 BC230804 NME16 p3ntsb08 BC230804 NME17 p3ntsb09 BC240804 NME18 p3ntsb10 BC240804 NME19 p3ntsb11 BC240804 NME20 p3ntsb12 BC250804 NME21 p3ntsb13 BC250804 NME22 p3ntsb14 BC250804 NME23 p3ntsb15 BC260804 NME24 p3ntsb16 BC260804 NME25 p3ntsb17 BC260804 NME26 p3ntsb18 BC270804 NME27 p3ntsb19 BC270804 NME28 p3ntsb20 BC270804 6/11/04 6/11/04 6/22/04 6/24/04 6/25/04 6/25/04 6/28/04 6/29/04 6/30/04 8/18/04 8/19/04 8/19/04 8/19/04 8/23/04 8/23/04 8/23/04. 8/24/04 8/24/04 8/24/04 8/25/04 8/25/04 8/25/04 8/26/04 8/26/04 8/26/04 8/27/04 8/27/04 8/27/04 Note: 1) o = 16.6 m2/g; collection spot area = 0.5 cm2 Sample Duration 12:00-12:43 14:03-14:55 13:21-14:13 13:03-13:49 12:05-12:57 13:28-14:03 13:04-14:00 12:17-12:58 12:24-13:04 14:48-15:30 10:46-11:54 14:18-15:15 15:55-16:13 12:16-13:11 15:27-16:07 16:42-17:00 11:58-13:03 13:28-14:28 15:18-15:47 12:23-13:19 14:18-15:20 15:55-16:05 12:40-13:35 14:05-15:00 15:44-16:01 12:30-13:30 14:25-15:08 15:52-15:57 Initial ATN 11.834 9.422 3.133 6.011 4.722 0.889 1.593 0.127 4.544 19.378 0.933 5.285 0.123 9.512 16.256 22.01 15.191 22.563 27.949 14.003 21.113 24.901 5.166 15.799 7.045 12.202 21.057 5.258 Total Volume 350.8 350.1 351.5 350.1 347.9 348.7 348.4 348.8 347.3 348.6 348.7 349.4 350.6 351.8 352.6 351.1 356.3 357.1 356.4 357.8 356.1 356.0 353.3 353.3 352.1 350.8 350.8 350.0 Series Flow Rate LPM) 4.85 4.85 4.85 4.84 4.85 4.85 4.77 4.85 4.78 4.85 4.84 4.76 4.72 4.73 4.73 4.73 4.81 4.76 4.63 4.81 4.70 4.70 4.87 4.79 4.82 4.82 4.83 4.83 Sample Time (min) 72.3 72.1 72.5 72.4 71.7 71.9 73.1 71.9 72.7 71.9 72.0 73.4 74.3 74.4 74.5 74.3 74.1 75.0 77.0 74.4 75.7 75.7 72.5 73.8 73.0 72.8 72.7 72.5 BC Mass Increment (ng/min) 17.440 2.601 4.807 3.188 2.869 3.148 2.200 2.429 2.825 15.011 6.791 1.896 2.415 3.100 2.584 3.562 4.775 2.595 2.2808 4.0708 2.6092 3.0903 4.0451 2.7753 4.512 3.1301 2.3608 5.8132 Total BC Mass (nc 1260 188 348 231 206 226 161 175 205 1079 489 139 179 231 193 265 354 195 176 303 198 234 293 205 330 228 172 422 Normalized BC Mass (ug/£ n/a - n/a n/a n/a n/a n/a n/a n/a n/a 874.595 398.579 113.120 145.757 187.526 156.448 215.025 287.709 158.173 142.751 245.539 160.577 189.595 237.618 165.898 267.835 184.550 139.449 341.649 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items