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.

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

PARTICULATE M A T T E R MEASUREMENT 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 S U B M I T T E D IN PARTIAL F U L F I L M E N T OF T H E R E Q U I R E M E N T S FOR T H E D E G R E E O F  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 ( M E C H A N I C A L ENGINEERING)  T H E UNIVERSITY OF BRITISH COLUMBIA  April 2007 © T i m o t h y X i Wang, 2007  Abstract This study develops and demonstrates a particulate matter ( P M ) 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 background 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 experiments (using methane and methane/ethane) did not produce noticeable B C trends with combustion temperature, pressure, or equivalence ratio ( E Q R ) . Non-premixed experiments with a gaseous fuel (using m e t h a n e / D M E , 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 uncertainties. 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 repeatable global B C mass measurements from methane flames in a shock tube. A n y future work to build upon the current sampling system and methodology should be carefully approached.  i  ii  Table of Contents Abstract  .  Table of Contents .  ii  '. .  iii  List of Tables  vii  List of Figures  .  ix  List of Symbols and Abbreviations  xii  Acknowledgements  xv  1  Introduction  1  2  Literature Review  3  2.1  Introduction  3  2.2  Background  4  2.3  Soot Formation Mechanisms  5  2.4  Direct-Injection Natural Gas Engines . . .  2.5  Particulate Matter Sampling from Shock Tubes  9  2.5.1  Shock Tube Dynamics  9  2.5.2  Particulate Matter Sampling Instrumentation  2.5.3" In-Situ P M Experiments in the Shock Tube 2.5.4 2.6 3  Shock Tube Experiments of P M Emissions  Conclusions  .  8  12 .  14 15 17  D e v e l o p m e n t 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  24  3.4  Elimination of Pressure Disturbances  Contamination Control . . .  25  iii  Table of Contents  3.5  3.6  3.7 4  •  v  3.4.1  Contamination Detection  3.4.2  Pre-Experiment Control  3.4.3  Post-Experiment Control  34  3.4.4  Particle Identification  39  "  Particle Loss Minimization  26 28  . . .  45  3.5.1  Inside Shock Tube  .  3.5.2  Gas Venting Process  49  3.5.3  Sample Container  50  3.5.4  Container to Aethalometer  52  46  Instrument Detection  54  3.6.1  Operation Principle  55  3.6.2  Data Algorithm Modifications  57  3.6.3  Error and Uncertainty  61  Conclusions  •  Particulate Matter Sampling from Methane Flames  64 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  4.2.4  Error Analysis  4.3  4.4 5  j  -.  73 78  Non-Premixed Study 4.3.1  Apparatus  4.3.2  Procedure  4.3.3  Results and Discussion  4.3.4  Error Analysis  Conclusions  Conclusions and Recommendations  80 .  81 .  83 84  ;  92 '93 96  References  99  Appendices  104  Table of Contents A Background A.l  B  104  Soot Formation Mechanisms  104  A . 1.1  Soot Precursors . . .  104  A . 1.2  Particle Inception . .  105  A . 1.3  Particle Growth . . .  107  A . 1.4  Oxidation  109  A.2 Shock Tube Apparatus . . .  111  A.3 Previous Work  111  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 Throttling 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 C.4.1  S E M Analysis  C.4.2  T E M Analysis .•  D Particle Loss Minimization D. l  E  F  :  127 127 -  s  Inside Shock Tube  130 138 138  D.2 Gas Venting Process  140  D. 3 Sample Container  140  Aethalometer Data Analysis  141  E. l  Operation Principle . .  141  E.2  Data Algorithm Modifications . . .  144  E. 3 Instrument Maintenance  146  Aethalometer Flow Meter Correction and Calibration  155  F. l  Flow Rate Correction  155  Table of Contents  '  Direct Correction  156  F. 1.2  Manufacturer-Suggested Correction  157 158  G Premixed Experiment Data  H  j  F.l.l  F. 2 Experimental Calibration  G. l  v  165  Procedure  165  G.2 Results  165  G . 3 Error Analysis  166  G . 3.1  Driven Gas Composition  G.3.2  Experimental Temperature/Pressure  G.3.3  Black Carbon Mass  167 . . .  170 170  Non-Premixed Experiment Data  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 G r i d 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 G r i d from Blank Experiment  137  C. 6 T E M Analysis of Lacy Carbon G r i d 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  159  Calibrated Flow Rates ( L P M ) . ., vii  List of Tables F. 2 Calibration D a t a G. l  viii . .  . .  160  Premixed Methane Series Experimental Test Matrix  165  G.2 Premixed Methane/Ethane Series Experimental Test Matrix  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 M e t h a n e / D M E Series Experimental Test Matrix  179  H.2  Non-Premixed Methane Series Experimental Test Matrix  179  H.3 Non-Premixed Methane/Ethane Series Experimental Test Matrix  180  H.4  181  Non-Premixed M e t h a n e / D M E Series Conditions  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 M e t h a n e / D M E Series Aethalometer Conditions  185  H.9  186  Non-Premixed Methane Series Aethalometer Conditions  H.10 Non-Premixed Methane Series Aethalometer Conditions (Continued) H . l l Non-Premixed Methane/Ethane Series Aethalometer Conditions  . .  187 188  List of Figures 2.1  U.S. Heavy-Duty Diesel Engine Emissions Regulations [53]  .  5  2.2  Aggregates of Primary Soot Particles [13]  2.3  Summary of Soot Formation Stages [51]  2.4  Shock Tube Principle [19]  3.1  Shock Tube Apparatus Schematic [19]  3.2  Double Diaphragm Setup [21]  3.3  Shock Tube Facility  3.4  Light and Pressure Signals from Combustion Test (20 bar, 1200 K )  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  6 . . .  7 11  .  22 ;  22 23 . . .  29  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 3.16 Typical Black Carbon Sampling Results  58 .  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 M e t h a n e / D M E Series Injection Results  85  4.7  Non-Premixed M e t h a n e / D M E 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]  Ill  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 W i t h i n Experimental Region (40 bar)  118  B.5  Diaphragm Fragment Trapping Devices  119  B. 6 Throttling Losses [19] C. l  Passing Through Shock Tube  C.2 Bypassing Shock Tube  '  '. .  ,  119 121  '  C.3 Blank Test Light Emission (20 bar, 1200 K )  122 .  123  C.4 Inert Blank Test Light Emission (20 bar, 1200 K )  123  C.5  124  Comparison of Tube Surfaces  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 C.22 Soot Particle 2  • • :  133 133  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 Fit  147  E.3  Polynomial Curve Fit (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 H. l  J-43 Mass Flow Correlation  .  167 178  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  ATN  Optical Attenuation Unit  bhp-hr  Brake Horsepower-Hour  BC  Black Carbon  cm  Centimeters  C  Carbon  C  Specific Heat at Constant Pressure .  CCD  Charge-Coupled Device  CMOS  Complementary Metal Oxide Semiconductor  CNG  Compressed Natural Gas  CPC  Condensation Particle Counter  d  Particle Diameter  D  Diffusion Coefficient  DMA  Differential Mobility Analyzer  DME  Dimethyl Ether  DMPS  Differential Mobility Particle Sizer  EDX  Energy Dispersive X-ray  EPA  Environmental Protection Agency  p  xii  List of Symbols and Abbreviations EQR  Equivalence Ratio Soot Volume Fraction  F dh a  Adhesive Force Thermophoeretic Force Grams  g H  Hydrogen  HPDI  High Pressure Direct Injection  ID  Internal Diameter  k  Specific Heat Ratio  L  Liters  LPM  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  PM  Particulate Matter  PM  1 0  Q rpm  Particulate Matter Under 10 (im Diameter Flow Rate Revolutions per Minute  R  Universal Gas Constant  RMS  Root Mean Squared  s  Seconds  xiii  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 appreciation 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 ( P M ) and oxides of nitrogen ( N O ) , vast amounts of research x  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 emissions while maintaining the inherent diesel engine benefits x  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 methodology, 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 effectiveness 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 ratio, 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 experimental 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 versus 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 P M 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 P M 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 P M 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 P M 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 oxidation 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 P M i o (under 10 /im in diameter) known for being carcinogenic, as well as causing respiratory problems. In addition, along with N O emissions, combustion P M contributes to smog x  and global warming. It is estimated that U.S. diesel powered vehicles produce 10  11  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 NO have dropped an order of magnitude from x  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 ( (ratio of soot volume to total volume), particle number density N (ratio of v  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 H [13]. Each primary particle is made 8  up of a large number (10 ) of crystallites, and contains up to 10 carbon atoms. The 4  6  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 fuelrich regions, and generally yields higher volume fractions than premixed flames. Temperature 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 temperature, species concentration, and the associated shifts in equilibrium. The final soot volume fraction increases with increasing pressure and increaing C / O ratio [51]. The critical 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 relative, 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  d-50nm Coagulation Surface growth and coagulation Particle inception Particle zone  —'  ••— d - 0 . 5 n m  f  Molecular zone  F i g u r e 2.3: Summary of Soot Formation Stages [51]  2.4. Direct-Injection Natural Gas Engines  2.4  8  Direct-Injection Natural Gas Engines  Development of natural gas (NG) fueled internal combustion engines has been accelerated 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 Innovations attempts to combine some of these approaches to achieve operational feasibility of N G injection, while realizing the l o w - P M producing nature of gaseous fuels. The H P D I 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, significantly 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 uncertainties 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 H P D I or other NG-fueled engine testing and research.  2.5  Particulate Matter Sampling from Shock Tubes  Particulate matter sampling is performed routinely from engine emissions whereby cylinK  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 complicated variable interactions and flow dynamics within the engine, it is rarely possible to change exactly one operating parameter while keeping all the other conditions constant. 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 H P D I 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. A t 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 equals 5  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 ) is the same as 5  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  Driven gas (a)  Diaphragm  11  Driver gas  Pi 3EZ  Incident shock front  Contact surface  (b)  Rarefacrionfan. ^—1~ -* U 4  Reflected shock  (c)  -flfront  1*  P  5  "5  Reflected rarefaction fan 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 undertailored 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. B y 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  2.5.2  12  Particulate Matter Sampling Instrumentation  Conventional engine emissions P M sampling usually involves direct mass measurements, such as the Tapered Element Oscillating Microbalance ( T E O M ) . If the P M mass is sufficient (tens of micrograms), fine collection filters are placed i n 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 structural 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 Mobility Particle Sizer ( D M P S ) is another popular particle sizing instrument in aerosol research. D M P S uses the principle of separating particle sizes by their electrical mobility. B y applying specific electric field strengths in a Differential Mobility Analyzer ( D M A ) , the target-sized particles will be forced into a secondary flow stream [31]. A Condensation Nucleus Counter ( C N C ) 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 absorbing 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 sample. 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 nonabsorbing 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  2.5.3  14  In-Situ P M 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 Williams [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 Wittig [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 extinction/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 P M 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~ ), global particulate matter sampling from methane 10  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 ( C N G ) 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 S O F 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 m g / 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 estimate 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 control. The relatively large quantities of P M generated (using liquid fuel sprays and premixed 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 mechanisms. The particle sampling paths and processes were also briefly mentioned without any elaboration of apparatus and/or procedural developments. In addition, most authors 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 sequence of stages: precursor formation, particle inception, particle growth, and oxidation. The fundamental chemical and physical processes are both kinetically and thermodynamically controlled, as well as being highly dependent on initial conditions. Numerical predictive modeling of soot formation is usually inadequate and thus experimental studies provide the bulk of the reliable data. To realize the P M emissions reduction potential of NG-fueled compression-ignition engines, Westport's H P D I uses a diesel pilot to ignite the main N G charge. The P M contributions from multiple sources are difficult to separate 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 challenge, 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 challenge, 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 TO  - 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, consideration 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. A s 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 particular 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 problem, 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 appropriate 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 , gas chromatography analyzers) x  and particle analysis instruments (e.g. particle sizing, number concentrations). If possible, 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, equivalence 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 Sensor  |-f V Vacuum Pump  Fuel  Driven Section  n  6  Static Pressure Sensor  1  Vacuum Sensor  yy^  Helium  Driver Section  F i g u r e 3.1: Shock Tube Apparatus Schematic [19]  F i g u r e 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 measurement device. Although stressed stainless steel particles are slightly magnetic, carbon steel shim diaphragms are used to ensure maximum magnetic capture efficiency. T h e 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 temperature 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 various 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 pressure 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 contamination 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 foreign 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 procedures. 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 immediately 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 contamination, 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.  T a b l e 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 pressure 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. A s 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 depend 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)  F i g u r e 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 contamination and post-experiment control methods. Cleaning Methods 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 striking 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) \ " drill, 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. B y 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 experiment, 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 drill bit' size constraints. Pressure ratios and axial speeds of approximately 5 and 15 c m / 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 connected 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) result 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. Particle Entry Prevention To complement the various cleaning methods developed, efforts to prevent particle entry into need to be taken to achieve thorough pre-experiment contamination control. The importance of ensuring that the post-cleaning environment is uncontaminated  3.4. Contamination Control  34  T a b l e 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 (H 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 2  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 postcleaning 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 important 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  Post-Experiment Control  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 preexperiment 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 Filtering 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 particlecontaining 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 d  n  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. Particle Settling Due to the inconsistent results from the various impactor designs, a more reliable diaphragm 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 experiment 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 experiment 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. Bag Contamination 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  T a b l e 3.3: Summary of Post-Experiment Control Procedures Post-Experiment Control Procedure Blank Test Summary BC Mass (ng) Temperature (K) Pressure (bar) Without Magnetic Filtering 795 - 835 1020 40 With Magnetic Filtering 420 - 460 1020 40 . Without Impactor Conventional Low-Pressure Impactor 1/2" Single-Stage Impactor Multi-Stage Impactor 1" Multi-Nozzle Impactor Settling Time: 0 min. vs 60 min. Settling Time: 10 min. vs 70 min. Settling Time: 0 min. vs 130 min.  940 - 980 935 - 975 345 - 385 670-710 330 - 370  1350 1350 1300 1300 1300  20 18 30 30 30  575 vs 95 315 vs 135 715 vs 90  1300 1300 1300  30 30 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  Particle Identification  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 ( S E M ) or the transmission electron microscope ( T E M ) to pick out individual particles. Detailed elemental analysis using energy dispersive x-ray ( E D X ) 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 Analysis 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  xlhO  F i g u r e 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 carbon 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 compositions 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 correction 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  E lem ent Carbon Oxygen Sodium Magnesium Alum inum Silicon Phosphorus Chlorine Potassium Calcium Iron  Concentration 24.86 wt% 53.50 wt% 0.73 wt% 0.35 wt% 3.58 wt% 14.62 wt% 0.30 wt% 0.11 wt% 0.16 wt% 0.16 wt% 1.64 wt%  10-  K Ca  Fe  -4-  Fe  keV  F i g u r e 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 TEM  Analysis  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 elemental 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 ent Carbon Oxygen Sodium Magnesium Alum inum Silicon Sulfur Chlorine Calcium Chromium Manganese Iron Nickel  0  2  4  6  Concentration 12.01 wt% 38.16 wt% 0.86 wt% 0.34 wt% 0.48 wt% 21.05 wt% 0.27 wt% 0.12 wt% 0.06 wt% 4.61 wt% 0.46 wt% 19.37 wt% 2.20 wt%  8  keV  F i g u r e 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 particles, 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 component 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, representative 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 contamination 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 i g u r e 3.10: Lexan Particle (1 /xm)  3.5  Particle Loss Minimization  While contamination control was primarily concerned with blank experiments and reducing 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 analyzed 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  Inside 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 continually 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. Thermophoresis 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.  -p\d?VT f VtH  (3.2)  -0.55r?Vr  (3.3)  PgT  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 F h and V / , under shock tube conditions. Since t  t  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. Settling 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 _  V V  pdg 2  P  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 ~ to 7 x l 0 5  - 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. Diffusion 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 diffusion, 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 transport, 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 §  '•  < > 35  The flux of soot particles is directly proportional to the diffusion coefficient, with sample calculations using typical soot -sizes shown in Appendix D. It can be seen that the l  number of particles diffusing through a unit cross-section per second in a unit concentration 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 aggregates. In addition, sub-micron particles with non-negligible slip factors will diffuse even faster with decreasing size (D proportional to d~ ). For example, a 50 nm particle will 2  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 g a s / P M 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  Sample Container  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 g a s / P M 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. Development 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 ) with ambient light. As a result, two different materials with the above properties x  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 ionization 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 material. 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  •Non-Conductive  20  Time (minutes) Conductive (Al layer)  30  40  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 will 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. B y 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  C - conductive, N - non-conductive  40 E" 35 E  2 I | o  30  •< 25 20  ><  c  3>  N  — :  •<  «.  s- «s 0  <=—=>  > N <=  in  £ 10  N  •  c  5»• —  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 comparable 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  50 45  -  54  A  L - long, S - short  g- 40  1,35 c  r 30  •«?  =>  L  •  «s  S  =>| «s L  c  E 25  >• • • • <  S  • ••  =>  o 20  8  15  5  10  to  •  e  L  => •  ->i  • •  5  0  <  • c «—o  i  i—— 10  1 —i 15 201 25 Time (minutes)  — * 1 _ 30 35  40  F i g u r e 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 interpretation 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 consistently 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 dedicated 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 collection 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 methods, 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 P M / g a s 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)  j  r  ATN = - 1 0 0 x ln(  a =  T  ~  )  (3.6)  Io — do  AATN  .  .  (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 I , respectively) are 0  measured. The factor of 100 is used for numerical convenience. B y 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. A t 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]. A t 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 n g / m at typical settings of 4 L P M flow rates and 1-minute timebase 3  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 g u r e 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. B y 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 A T N 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 U V 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 -10  • " •  E  1  8  E  6  •  •  •  •  •  •  •  s>  o  in in  4  10  15 Time (minutes)  20  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 gravitational 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 significant 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  ABC  •c* 20 E ~5) c  r  >UV  * • • • ,  —  15  c a> E a> b 10 _c in in «5  r-  : A A  A  AAAAAA AA AA * A  A  — AA AAA  A  A  •  AAAAAA AA AAAAAAAA AA*  A  A  A  A  A I  0  :  A  10  ;  |  20  1  30  .  1  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 wavelengths (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 estimate 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 characteristics of different types of contamination particles under U V light are needed, as well  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 direct 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 documented 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 nonuniform 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 manufacturer calibration is a combination of absorption and scattering contributions [40], where the variations in the scattering a component can result in overall a values between 8 m / g and 19 m / g . Based on the testing location conditions, the difference in 2  2  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 / g is obtained from an 2  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. Experimental 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  3.7  64  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 consistency, 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 diaphragms 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 emissions, 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 electrostatics, 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 P M / g a s 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 existence. 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 experiment, provided that it is sufficiently above the baseline levels. The premixed and non-premixed methane control experiments have been helpful to locate the critical experimental 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 are notorious diesel engine pollutants. However, the combustion variable X  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. A n isolation facility such as the shock tube, therefore, is crucial for studying P M emissions, especially in conjunction with the H P D I 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 investigate 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 injected 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 discrepancies 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 preliminary 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 comparison 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 experimental 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 will 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 i g u r e 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 p s 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 p s 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 contamination 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 / g and a measured tape spot area of 0.5 2  c m . These mass increments are then plotted against the time lapse from the sampling 2  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. T h e 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 temperature 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  .  ——A  100  A  :  j  —  3  80  O  CO  "a a) N 15  •  60  A  A EQR<1 » EQR>1  40  E L_  o  A  9  20  "  1020  • EQR=1  1  1070  1  1120  •  1  1  1220  1270  1—•  1170  Temperature (K)  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 ( E Q R < 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 M 3 to M 9 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 ( E Q R > 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 40 in  EQR=1 A EQR=0.5 • EQR=2 [xEQR=3 + EQR>3  TO  5 O  m 30 T3 0)  N  15 20  E-  10 X 1000  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 i g u r e 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 M E 8 to M E 1 2 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 M E 1 2 data. However, once this sooty condition is reached, it is very difficult to restore the tube to its 'clean' state, where the subsequent E Q R = 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  V  X  25 D)  3  20  • 15 +/-1 bar  + *  O m  -u 15 <u N re  A  X  «41 +/- 3 bar  x  E i_ 10 o  25 +/- 3 bar  x Methane  •  -tx  + Methane/Ethane  x  x  •  X  -I  0 1000  ±  ,  1050  *  +  *  ,  1100 Temperature (K)  , 1150  X  ++ x 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. A l 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 ( B l , 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 contamination. 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 (especially 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 complex. 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 cleaning 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  4.2.4  78  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. Driven Gas Composition 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 . Experimental Temperature/Pressure The experimental temperatures and pressures are calculated from the driven gas composition 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. Black C a r b o n Mass Total black carbon masses from each experiment are calculated from the most representative 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 measurement 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 deviations 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  Premixed Methane  80  T a b l e 4.1: Comparison of Standard and Measurement Errors' Average Standard Average Measurement Errors Normalized Number Standard Error Fuel Normalized BC Mass of Expts Deviation • aUfN Mass (%) BC Mass (%) (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 H P D I technology) requires non-premixed investigations of natural gas-like fuels. These non-premixed experiments also allows an approach in the above technical challenges 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 ( D M E ) 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 procedures 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  Apparatus  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. A n 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 m e t h a n e / D M E 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 m e t h a n e / D M E 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 injector 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 W C u t 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. Pressurized standard Praxair fuel mixtures of m e t h a n e / D M E (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 mentioned 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 monitoring 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 correlation 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 W C u t 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 m e t h a n e / D M E 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  Results and Discussion  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.  120  • 8+/-1 bar A  100 in m O CO TJ  80  A11+/-1 bar  • 17+/-1 bar x 23+/-2 bar X  X  •  60  X  a>  N  40 •  20  m r—  800  1000  1  —  1  1200 1400 Temperature (K)  —  , 1600  1800  F i g u r e 4.6: Non-Premixed M e t h a n e / D M E 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  1300 1100 CO  in in  ro O CO XI V N  900 700  86  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  500  —  -«-  + —  o  + 300  X  •  V  x  +  o t  800  1000  o  i  +  100  o  1200 1400 Temperature (K)  , : 1600  1800  Note: Prefix B indicates Blank Series F i g u r e 4.7: Non-Premixed M e t h a n e / D M E 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 N M D 1 7 B and N M D 1 8 B , 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  350  ~  87  ~  ~~  5  •'  ;  300 f 250  :  .  O  to in  n3 5  O  0  200 o  o  3 150 N  0  § 100 50  ;  0 ' 1100  '.  O  o  \* 1  1150  1  ,  1200  1  '  1  1250 1300 Temperature (K)  1  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 F i g u r e 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  -O  —  o Injection A Blank  _ 2100  X  D) C  tn  co  _  o  1600  o  o  0  A  CO  JCO -1100 T3  A*  ©  I  * A  E  600 0  100 1100  _! 1150  *A A  ©  1200  i  1  0  •—i  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 i g u r e 4.9:.Non-Premixed Methane Series Blank Results polyolefin bag (in addition to the advanced contamination control and cleaning procedures 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 experiments. 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  400  89  ~~  —  ©  _ 350  —  ©  D)  300 co  © o  O 250 •o CD N  H  -  ©  © ©  200  0  o  150 100 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: 875 ug/g Figure 4.10: Non-Premixed Methane/Ethane Series Injection Results impactor (starting at N M 3 6 B ) proved ineffective, where both the B C masses and linear curve decay returned to pre-impaction levels (also confirmed by N M 4 5 B ) . 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 will 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  ©  o Injection A Blank  450  1?  ©  400 350  © ©  O oo 300  -o o> .b! 250 "fo E o 200  ©  o ©  A A A A  ©  A  L_  A  150  A  © ©  ©  ©  %  © ©  ©  100 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 i g u r e 4 . 1 1 : 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  1700 A_A  1500  in in co  1100  O CD  900  73 0)  N "CO  E t~ o  J  A  cn 1300  X  o  A  o  -  0  700  o  500  — o  300  o Methane/DME A Methane x Methane/Ethane  ——  100 1100  &-  .  /.  • • o  A AX y\—rt——  ** • . 1  1200  *  9  -  :  « —  o ______  o o  1  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 N M E 3 B ) 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 issues 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 comparisons 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 conditions 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 combustion 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 will also be ignored at this preliminary stage. Errors in fuel compositions will 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 m e t h a n e / D M E 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. A s 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 m e t h a n e / D M E , 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 methodology 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 quantitative 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 interpretation. 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 thermopheoretic 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 demonstrate 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 nonpremixed 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 natural 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 and gas x  chromatography studies using this sampling system can provide complementary understanding 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-nheptane 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 summertime hourly black carbon measurements in southwestern pa. Atmospheric Environment, 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, 1 0 7 ( A A C ) : X 1 - X 9 , 2002. [5] K. Brezinsky. High Pressure Shock Tube Gas Sampling. University 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  REFERENCES [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-Hill, New York, 1988. [16] W . C . Hinds. Aerosol Technology: Properties, Behaviour, and Measurement of Airborne 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 British 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 enginerelevant conditions. Master's thesis, University of British 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 Manual. Monterey, C A , 1994.  Instruction  [24] T . P . Jenkins and R . K . Hanson. Soot pyrometry using modulated tion/emission. Combustion and Flame, 126:1669-1679, 2001.  absorp-  [25] H. Jones. Particulate matter emission sources in hpdi engines and strategies to reduce pm emissions. Master's thesis, University of British Columbia, 2004.  REFERENCES  101  [26] H. Jung, D . B . Kittelson, and M . R . Zachariah. Kinetics and visualization of soot oxidation using transmission electron microscopy. Combustion and Flame, 136:445456, 2004. [27] H. Kellerer, R. Koch, and S. Wittig. 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. Wittig. Soot formation in a shock tube under elevated pressure conditions. Combustion Science & Technology, 113/114:6780, 1996. [29] I.M. Kennedy. Models of soot formation and oxidation. Combustion Science, 23:95-132, 1997.  Progress in Energy and  [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: Apparatus, theory, and applications. Journal of Aerosol Science, 6:443-451, 1975. [32] C. Kopp, A . Petzold, and R. Niessner. Investigation of the specific attenuation crosssection 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 measurements of black carbon in an occupied house: A n 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; implications on the calibration curve of an inertial impactor. Journal of Aerosol Science, 31:923-932, 2000. [37] A . Muller and S. Wittig. 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:  REFERENCES [38] B.R. Munson, D.F. Young, and T . H . Okiishi. Fundamentals of Fluid John Wiley & Sons, third edition, New York, 1998.  102 Mechanics.  [39] C. Park and J.P. Appleton. Shock tube measurements of soot oxidation rates. Combustion 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 Environment, 31(5):661-672, 1997. [41] S.N. Rogak. Shock Tube Cleaning Notes. University of British 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, 26:324334, 1999. [46] R . E . Sonntag, C . Borgnakke, and G . J . Van Wylen. Fundamentals of Thermodynamics. 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 i n , 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 different hydrocarbons behind reflected shock waves. Combustion and Flame, 112:359370, 1998. [51] J . Warnatz, U. Maas, and R . W . Dibble. Combustion: Physical and Chemical Fundamentals, Modeling and Simulation, Experiments, Pollutant Formation. SpringerVerlag, third edition, New York, 2001.  REFERENCES  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 hydrocarbons such as C i , C 2 , and in particular C H 2  2  (acetylene). The initial step in the production  of soot is the formation of the first aromatic species from these smaller hydrocarbon fragments. 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 ( P A H ) [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 with 2  C2H2  to form  After recombination to an aliphatic and rearrangement, the first ring (benzene:  C3H3.  ) is  formed as shown in Figure A . l , partly aided by the very slow competing oxidation reactions of  C3H3.  A n example of the elementary reaction mechanism for P A H growth from  acetylene starts with the addition of C H 2  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 oxidized 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 g u r e A . l : Precursor Formation Process [51]  A. 1.2  Particle Inception  The production of soot particles in a flame is inherently a chemically controlled phenomenon. Thermodynamics alone cannot describe this process since soot is formed beyond 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 as the main 2  products) can be represented by: 777  CH n  m  + k0  2  -  2kCO + -H  2  + ( n  2k)C  s  (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 formation 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 conglomeration 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 particles (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. A t 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 considerable experimental support. A t 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. Experiments 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 hydrocarbon 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 hydrocarbon 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)  J  ^H, CH,  Sool  Indirect (slow)  Soot  Aliphatics  F i g u r e 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 P A H s 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, f is related to N and d by v  (A.3)  f = (^)Nd  3  v  The rate of change of particle number density with time can be expressed as dN  = N  n  where N  n  - N  (AA)  a  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 ) , the particle diameter remains essentially constant at the n  a  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 ) , the number of n  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. W i t h 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 g u r e 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. Oxidation in this case also continues downstream as long as the temperature is high enough. Although thermodynamics alone cannot adequately describe the soot formation phenomenon, 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: CH 6  5  + CH 2  2  «-» (C H C H ) 6  5  2  2  <-+ C H C H e  5  2  + 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]. A n 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 hypothesis 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.  A.2  Shock Tube Apparatus  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.  Injector Nozzle^ Quartz Window F i g u r e A . 4 : Injector Connection Setup  A.3  [21]  Previous Work  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.  •  . Muinor\s;  i  . rear  Alexiou, Williams  1995  Alexiou, Williams  1996  Cadman  1996  S u m m a r y of In-Flame P M E x p e r i m e n t s Reaction Pressure Temperature (bar) (K)  Fuel(s)  Measurement Quantity  toluene/n-heptane, toluene/iso-octane toluene/methanol, toluene/ethanol, etc. ethanol  Pyrolysis  0.19-0.36  1600-2400  Pyrolysis  1.7-3.5  1450-2450  Injection  12  1000-1800  soot induction time, formation rate soot induction time, formation rate soot yield  Cadman, Denning 1996  oxygen  Premixed  4-14  1500-3500  soot oxidation rate  oadman, Denning 1997  hydrogen/oxygen  Premixed  3-15  1150-3000  soot oxidation rate  toluene, benzene, ethylene, methane, etc. methane, ethylene, acetylene, propane, etc. propanol/tetradecane/ water methane, ethylene, acetylene oxygen  Premixed  10-60  1500-2300  Premixed  15-100  1600-2100  Injection  4-9  1000-2200  soot particle size, number density, volume fraction soot particle size, number density, volume fraction soot yield  Premixed  4-9  1600-2000  Premixed  0.05-13  1700-4000  Kellerer, Wittig  2000  Kellerer, Muller  1996  Kunz, Cadman  1997  Muller, Wittig  1991  Park, Appleton  1973  — Muuioqsj  rear  Sidhu, Graham  2001  Wang, Cadman  1995  Wang, Cadman  1998  \ Fuel(s)  soot induction time, particle size, concentration soot oxidation rate  Summary of P M Emissions Experiments Reaction Pressure Temperature (bar)  (K)  CNG, DME, biodiesel, diesel benzene  Premixed Injection Injection  20-27  1000-1500  2  2400  toluene, n-heptane, propanol-1  Injection  2-25  1000-3000  Measurement Quantity  particle size, shape, SOF C  60  soot yield, PAH yield  Measurement Method  laser beam attenuation 632.8 nm laser beam attenuation 632 8 1152 0 nm laser absorption 488 nm laesr beam attenuation 488 633 nm laser absorption 488 nm laser light scattering 488 w 632 8 nm V^iU M i l l laser light scattering 488 632 • w . 8 i nm n if laser absorption 488 nm optical dispersion quotient; 488, 633 nm laser-light transmission fi39 R nm  Measurement Method  filter, gravimetic, SEM, thermal desorption filter, mass spectrometry filter, gravimetic  A.3. Previous Work  '^  ;  113  Manual Valve Solenoid Valve  Quartz Windows (3)  Optical Test Section  Injector  Pressure Transducers Adaptor  ^ Clamping Plates Rubber Instrumentation Window  F i g u r e A . 5 : Optical Access Section [21]  LShock tube wall  V Ignition inside the shock tube  .Quartz window  Optical fiber  Bandwidth filter Photo multiplier  J\  Data acquisition system  o o o  F i g u r e A . 6 : Optical Access Principle [19]  r V  A.3. Previous Work 114  High Pressure Sbocktube Croup Sampling  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 preshock 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 ricjincrrincj  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 performance 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 representative 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 pressures 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. A n y 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 diaphragms 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 devices, 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 ). After closing and fastening the shock tube, 0  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 - V ) Q  B.4  (B.l)  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 tailoring, 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 L  1  1  1.5  F^W^.  ^  W  u  ,  0.5 i  i _ .  .  15  Time (ms)  F i g u r e B . 4 : Pressure Disturbances Within Experimental Region (40 bar)  n — (polytropic ratio) x k  PV  n  = constant = P{V? = P Vf 2  (B.2)  (B.3)  B y decreasing the polytropic ratio from the isentropic value of one, the expansion process 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 i g u r e B . 6 : Throttling Losses [19]  B.4. Determination of Throttling Losses  120  T a b l e B . l : Summary of Burst Pressure Testing  Material  Thickness  Brass Shim  0.003"  Stainless Steel Shim Stainless Steel Shim . Stainless Steel Shim Stainless Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Transparency Film Lexan (matte) Lexan (clear)  0.001" 0.002" 0.003" 0.004" 0.002" 0.003" 0.004" 0.006" 0.007" 0.008" 0.003" 0.010" 0.010"  Lexan (clear)  0.020"  Lexan (clear)  0.030"  Transducer Voltage (V) Atmospheric Burst  0.451 0.451 0.448 0.447  0.526 0.522 0.542 0.63  0.451 0.441 0.448 0.448 0.45 0.452  1.129 0.507 0.549 0.587 0.65 0.768  0.451 0.451 0.447 0.447 0.447 0.447 0.451 0.451 0.451  0.551 0.53 0.549 0.549 0.635 0.644 0.765 0.739 0.747  Burst Pressure (psi)  * Note: estimated from pressure regulator gauge  T a b l e B . 2 : Summary of Diaphragm Polytropic Ratios Material  Brass Shim Stainless Steel Shim Stainless Steel Shim Stainless Steel Shim Stainless Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Carbon Steel Shim Transparency Film Lexan (matte) Lexan (clear) Lexan (clear) Lexan (clear)  Thickness  0.003" 0.001" 0.002" 0.003" 0.004" 0.002" 0.003" 0.004" 0.006" 0.007" 0.008" 0.003" 0.010" 0.010" 0.020" 0.030"  Polytropic Ratio  0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.95 0.9 0.9 0.9 0.9  94 89 118 229 450 * 848 83 126 174 250 395 450* 125 99 128 128 235 246 393 360 370  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 i g u r e C . l : Passing Through Shock Tube  121  C.2. Pre-Experiment Control  122  4  ass 1ncrement (ng/min)  3 5 -  _>  32 52 i >  1 51-  0C  i  |  2  1  4 6 Time (minutes)  Figure C . 2 :  C.2  <  1  8  10  Bypassing Shock Tube  Pre-Experiment Control Table C . l :  Summary of Total Particle. Counts  Ambient Lab Air Filtered Bottled Air, Through Tube Filtered Bottled Air, Bypassing Tube  C.3  •  •  0.5 -  Particle Count (#/ccm) > 50000 -50 ~ 2  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(5 —) o  (C.l)  C.3. Post-Experiment Control  0.1, 0.09 0.08 0.07 r-  I 0.06 TS  c  0.05  g>  w 0.04  Z  gi  - 0.03 1  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  Figure  Clean 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  Figure C . 9 : Magnetic Filter  125  C.3. Post-Experiment Control  126  where L is the nozzle to collector plate distance, d is the nozzle diameter (as shown in n  Figure C.10), and S is the particle Stokes number at stagnation conditions upstream of 0  the nozzle. After substituting the expression for S and assuming 0  m ~d p  where m  p  (C.2)  3 a  is the particle mass and d is the aerodynamic diameter. The linear expression a  of St with the particle diameter is  st = 2m ? '; ' ° "r) 0M8  f  d c  7  (c.8).  3  where P is the gas stagnation pressure, p is the particle density, c is the carrier gas 0  p  a  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 efficiency as a function of ^ exhibits a step function resemblance for hypersonic impactors.  Po dn i p  1  L  / I  <  Figure C . 1 0 : 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  impactor # of holes 1 3 2 5 3 10  128  d„ 1/4" 3/8" 5/8"  L 1/8" 3/16" 5/16" impaction  bleed flow  F i g u r e C . 1 2 : Multi-Stage Impactor Schematic  F i g u r e C . 1 3 : Single Stage Multi-Nozzle Impactor: Nozzle and Impaction Plates  CA. Particle Identification  Figure C.14: Dust Particle  kCts  15H  104  54  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%  Na  AA  t?  -i  Figure C.15: Dust Composition  1—i  8  r  keV  C.4. Particle Identification  T a b l e 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 i g u r e C . 1 6 : T E M Sampling Setup  C.4. Particle Identification  F i g u r e C.18:  131  Lexan Particle (100 nm)  C.4. Particle Identification  F i g u r e C.20: Lexan Particle (150 nm)  132  C.4. Particle Identification  Counts  200  -f  -J  F i g u r e C.22: Soot Particle 2  CA. Particle Identification  F i g u r e C . 2 3 : Soot Particle 3  F i g u r e C . 2 4 : Soot Composition  CA. Particle Identification  F i g u r e C . 2 6 : T E M Soot Photograph 2  135  C.4. Particle Identification  136  T a b l e C.3: S E M Analysis of Lacy Carbon Grid from Blank Experiment Elemental Composition (Approx. %) Size (um) Shape Material Background 1 C, O (44), Cu (45), Si (10) Background 2 C, O (56), Cu (43) 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, 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 1 0.5 1 0.5 1.5 0.5 2 2 3 8 0.5  oval oval rectangular needle circular  Lexan Lexan Lexan Lexan Lexan dust Lexan dust  Particle Density ~ 20-30 particles per square grid (40 um X 40 um)  T a b l e 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  T a b l e 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 Particle 2 Particle 3 Particle 4 Particle 5 Particle 6 Particle 7 Particle 8 Particle 9 Particle 10 Particle 11  C, Cu C, 0, Si, Cu C, 0, Si, Cu C, 0, Si, Cu C, Cu C, 0, Si, Cu C, 0, Si, Cu C, 0, Si, Cu C, 0, Si, Fe, Cu C, O, Si, Cu C, 0, Si, Cu  100 100 100 100 100 50 100 100 100 100 100  Soot Lexan Lexan Lexan Soot Lexan Lexan Lexan Lexan Lexan  T a b l e 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. Ad F a d h  =  (D.l)  12x  where A is a material-dependent constant, d is the particle diameter, and x is the separation distance (based on the surface roughness). Typical shock tube relevant parameters result in (75 x 1CT F  ™* = -  20  J) x (500 x 1 0 - m)  ,  9  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. -p\d?VT  F = -=—=  ~T  th  (D.3)  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 Pg  1  where rj and p are the gas viscosity and density, respectively. g  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 proportions, F h and V h values are estimated as t  t  - ( 1 x 10 Pa)(0.16 x 10~ m)(500 x 1 0 ' m) (30000 K/m) „ Fth = — ^ '-> ~ - 4 x 1 0 - TV 5  6  9  2  (D.5) -0.55(198 x 1 0 - /Vs/m )(30000 K/m) , (0.262 k,/J)WK) ' ~ 0-0042 m / , 7  . (D.6)  2  n  n  / r v  Settling velocities of.sample soot particles are shown below.  where p is the particle density and g is the acceleration due to gravity. p  V V  t  S  V V  t  S  (  2  5  Q  0  ^ /  ~ ~  m  3  ) (  5  0  X  1  0  "  9 m  ) ( ' 2  9  0*m-TmU  8  18(198 x 1 0 - ' Ns/m>) ( 2 5  2  ° ° W m ) ( l x l 0 - « m) (9.8 m / s ) 18(198 x 1 0 - ' Na/m') 3  2  X  l  °  7  v  7  X  i n 1  /S  _  2  (TSR\  ™  0  ( D  «  5 m  /  s  8 )  J  m (  "  D  '  9  )  The diffusion coefficients for sample 50 nm and 500 nm diameter soot particles are calculated below.  where k is Boltzmann's constant and C is the slip correction factor (dependent on particle size). c  '  "_  (1.38 x 1 0 -  2 3  (1.38 x 1 Q -  23  J/K){300 A")(1.3)  r  2  J/AT) (300 AT) (5.0) 2  2  2X  1  0  m  ,  ,  n  ,  q  ^ ~ 3,(198 x 1 0 - i V s / m ) ( 5 0 x 10~* nm) ~ " 7  o  '  s  (D-12)  .  D.2. Gas Venting Process  140  T a b l e 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 .10 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: F  E  =K ^ B  (D.13)  where q and q' represent the charge magnitudes, R is the separation distance, and K  B  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 = - 1 0 0 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 7 (A) = 71(A) x F(X) x OC(X) x D(\) 0  (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 / .= 7 (A) x e~  (E.3) .  AW  0  where the absorbance due to a particular aerosol species is  A(X) = k(\) x M A  P  (E.4)  and M p 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 ). This 2  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 M  .(E.5)  BC  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. T a b l e E . l : Summary of Specific Attenuation Values Channel Infrared (IR)  Wavelenqth 880 nm  °" (Black Carbon) 16.6  12.6  Ultraviolet (UV)  370 nm  39 5  30.0  cr (Elemental Carbon)  Intercomparisons with other analytical techniques were used to validate the Aethalometer'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 Oxidation, 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 carbon is interpreted as a subset of the elemental carbon mass. The second channel on the Aethalometer illuminates in the near-ultraviolet ( U V ) , 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*)xC(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 M  (E.7)  P  A  while the 370 nm wavelength U V channel measurement yields an absorbance of  A(X*) = k(^)  x (M  BC  +M  )  UVPM  (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.  ATN  2  - ATN, = (-100 x (In 3-0161 - 0-0196 3.2907 - 0.0191 V  V  _ ^  ;;  3.0213 - 0.0196 3.2948 - 0 . 0 1 9 1  ( k  = 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 ATN  3 . 7 5 9 - 3.713  — ATNi  2  d(BC) =  - =  a M  BC  = d(BC) x A  , (E.10)  2  2  16.6 m /g z  = 0.00277 g/m x (0.5 cm ) x (1 x 10~ m /cm ) = 1.385 x 1 0 " g 2  2  spot  „„„„„„ . . = 0.00277 g m  4  2  7  2  (E.ll)  \ = /f  [BC] — — - — ~ B ^ T T T o ^ (Flow Rate)(Time)  85  * ~ g 1 0  (looo  7  x  (4 L/mm)(1 mm) =  v  L  /  m  /  3)  x  ( 1 Q  9  /  }  x  Q  Q  l  ^ 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 n g / m ) due to roundoff 3  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  is M  BC  =  ATN - ATN, 2  A  xA  spot  (E.13)  E.2. Data Algorithm Modifications  M  BC  =  2 602 — 2 319 2/ (°x  1  6 6  m  5c m 2  g  )  145  (  x  1 0 _ 4  m  V c m ) x (10 ng/g) x 0.01 = 8.524 ng 9  2  (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. M  =M  BC  ln( L, _  * "  X U  1 5 7 1 0 8  )  ln(  V 184.187'' _  -5  ~  1 2 6 4 1 6  )  lnf  " V 184.187 > _  -10  ~  .  x e~  kt  BCo  1 1 1  1 1 4 6 0 8  )  ln(  V 184.187/ _  -15  (E.15)  "  1 5 1 5 1  )  ' " V 184.187 >  -75  0  CW TR\  flQQ  n  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 characteristics 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, entrainment 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 volumeto-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 representative 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 Time (minutes)  80  100  F i g u r e 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 minimize 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)  F i g u r e E . 2 : Linear Curve Fit  F i g u r e E . 3 : Polynomial Curve Fit (Type 1)  40  E.3. Instrument Maintenance  Time  T a b l e E . 2 : Sample Black Carbon Channel Data [BC] (ng/m )  Flow Rate (LPM) 365 4 343 4 348 4 -10795 3.9 -2292 3.6 1689 3.6 2317 3.6 2556 3.6 2660 3.6 2645 3.6 2663 3.6 2777 3.6 2782 , 3.6 2790 3.6 2601 3.6 2768 3.6 2742 3.6 2697 3.6 2608 3.6 2698 3.6 2651 3.6 2610 3.6 2460 3.6 2495 3.6 2523 3.6 2506 3.6 2369 3.6 2361 3.6 2399 3.6 2272 3.6 2171 3.6 2260 3.6 2222 . 3.6 15259 4 941 3.9 518 3.9 424 3.9 J  11:42 11:43 11:44 11:45 11:46 11:47 11:48 11:49 11:50 11:51 11:52 11:53 11:54 11:55 11:56 11:57 11:58 11:59 12:00 12:01 12:02 12:03 12:04 12:05 12:06 12:07 12:08 12:09 12:10 12:11 12:12 12:13 12:14 12:15 12:16 12:17 12:18  148  Sensing Offset (d) 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196 0.0196  Sensing Signal (I) 3.0213 3.0161 3.0168 3.0625 3.0696 3.0627 3.0565 3.04/4 3.0334 3.0224 3.0148 3.0051 2.991 2.9801 2.9728 2.9632 2.9499 2.9391 2.9324 2.9231 2.9102 2.8998 2.8936 2.8852 2.873 2.8631 2.8573 2.8495 2.8379 2.829 2.8239 2.8169 2.8057 2.7444 2.7427 2.7416 2.7371  Reference Reference Offset (d ) Signal (l ) 0.0191 3.2948 0.0191 3.2907 0.0191 3.293 0.0191 3.2962 0.0191 3.2947 0.0191 3.294 0.0191 3.2966 0.0191 3.297 0.0191 3.2925 0.0191 3.2911 0.0191 3.2934 0.0191 3.294 0.0191 3.2895 0.0191 3.2886 0.0191 3.291 0.0191 3.2914 0.0191 3.2874 0.0191 3.2862 0.0191 3.289 0.0191 3.2893 0.0191 3.2853 0.0191 3.2839 0.0191 3.2866 0.0191 3.2869 0.0191 3.283 0.0191 3.2816 0.0191 3.2843 0.0191 3.2847 0.0191 3.2808 .0.0191 3.2794 0.0191 3.2821 0.0191 3.2828 0.0191 3.2785 0.0191 3.2721 0.0191 3.2741 0.0191 3.275 0.0191 3.2715 0  ATN  0  3.713. 3.759 3.806 2.392 2.112 2.319 2.602 2.914 3.239 3.562 3.887 4.226 4.565 4.906 5.224 5.561 5.895 6.224 6.542 6.87 7.192 7.51 7.808 8.111 8.418 8.722 9.009 9.295 9.586 9.861 10.124 10.398 10.666 12.693 12.818 12.887 12.943  1  E.3. Instrument Maintenance  149  T a b l e E . 3 : Attenuation Value Changes Due to Gas Composition Start of Sampling (Air to Helium) Time (min.) BC (ng/m ) ATN 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 a  End of Sampling (Helium to Air) Time (min.) ATN BC (ng/m ) 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 3  30  Time (minutes) F i g u r e E . 4 : Polynomial Curve F i t (Type 2)  E.3. Instrument Maintenance  150  T a b l e E . 4 : Sample Post-Processed Aethalometer. Data Time (minutes) 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  [BC] (ng/m )  ATN  a  1689 2317 2556 2660 2645 2663 2777 2782 2790 2601 2768 2742 2697 2608 2698 2651 2610 2460 2495 2523 2506 2369 2361 2399 2272 2171 2260 2222  2.319 2.602 2.914 3.239 3.562 3.887 4.226 4.565 4.906 5.224 5.561 5.895 6.224 6.542 6.87 7.192 7.51 7.808 8.111 8.418 8.722 9.009 9.295 9.586 9.861 10.124 10.398 10.666  Mass Increment (ng/min) 8.524096386 9.397590361 9.789156627 9.728915663 9.789156627 10.21084337 10.21084337 10.27108434 9.578313253 10.15060241 10.06024096 9.909638554 9.578313253 9.879518072 9.698795181 9.578313253 8.975903614 9.126506024 9 246987952 9.156626506 8.644578313 8.614457831 8.765060241 8.28313253 7.921686747 8.253012048 8.072289157 :  E.3. Instrument Maintenance  T a b l e E . 5 : Exponential Decay Test Data Time (minutes) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 75 76 77 78 79 80 81 82 83  [BC] (ng/m )  ATN  54216 52277 49891 47517 46657 44251 43439 42620 40897 39640 35195 35228 34320 33681 32462 31920 31369 31170 7090 5997 4742 4109 3319 2954 2290 1553 1074  15.894 22.303 28.418 34.241 39.958 45.378 50.697 55.913 60.918 65.768 70.072 74.38 78.577 82.694 86.66 90.558 94.389 98.194 6.431 7.161 7.74 8.243 8.651 9.018 9.306 9.505 9.645  J  Mass Increment (ng/min)  1  193.042 184.187 175.392 172.199 163.253 160.211 157.108 150.753 146.084 129.639 129.759 126.416 124.006 119.458 117.410 115.392 114.608 26.024 21.988 17.440 15.151 12.289 11.054 8.675 5.994 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 Time (minutes)  30  35  ,  40  F i g u r e E . 5 : Steady Mass Increment Curve  250  10 Time (minutes)  15  F i g u r e E . 6 : Sooty Experiment B C Mass Increment Curve  20  E.3. Instrument Maintenance  Time (minutes) 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  T a b l e E . 6 : Sample Post-Processed [BC] BC ATN BC Mass Increment < 9 ) (ng/min) 282 1.987 528. 2.052 1.957831325 629 2.13 2.34939759 756 2.223 2.801204819 778 2.319 2.891566265 768 2.414 2.861445783 811 2.514 3.012048193 837 2.617 3.102409639 820 2.718 3.042168675 718 2.807 2.680722892 830 2.909 3.072289157 790 3.006 2.921686747 720 3.095 2.680722892 783 3.191 2.891566265 754 3.284 2.801204819 867 3.391 3.222891566 832 3.493 3.072289157 1340 3.658 4.969879518 854 3.763 3.162650602 911 3.875 3.373493976 858 3.98 3.162650602 863 4.086 3.192771084 931 4.2 3.43373494 924 4.314 3.43373494 803 4.412 2.951807229 229 4.44 0.843373494 792 4.538 2.951807229 839 4.64 3.072289157 772 4.735 2.861445783 765 4.829 2.831325301 728 4.918 2.680722892 775 5.013 2.861445783 637 5.091 2.34939759 793 5.188 2.921686747 777 5.283 2.861445783 904 5.393 3.313253012 774 5.487 2.831325301 813 5.587 3.012048193 796 5.684 2.921686747 820 5.784 3.012048193 768 5.878 2.831325301 817 5.977 2.981927711 858 6.082 3.162650602 849 6.186 3.13253012 795 6.283 2.921686747 869 , 6.388 3.162650602 847 6.491 3.102409639 917 • 6.603 3.373493976 767 6.697 2.831325301 n  /mJ  153  B C and U V Channel Data [UVPM] (ng/m )  UV ATN  a  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  UVPM Mass Increment (ng/min) 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  Table Time (minutes) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20  E.7: Sooty Experiment B C Data [BC] ATN Mass Increment (ng/m3) (ng/min) • 18152 9.253 54216 15.894 200.0301205 52277 22.303 193.0421687 49891 28.418 184.186747 47517 34.241 175.3915663 46657 39.958 172.1987952 44251 45.378 163.253012 43439 50.697 160.2108434 42620 55.913 157.1084337 40897 60.918 150.753012 39640 65.768 146.0843373 35195 70.072 129.6385542 35228 74.38 129.7590361 34320 78.577 126.4156627 33681 82.694 124.0060241 32462 86.66 119.4578313 31920 90.558 117.4096386 31369 94.389 115.3915663 31170 98.194 114.6084337 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  F.l.l  156  Direct Correction  The flow rate of an actual gas (helium/air mixture i n this case) is related to that of the reference/calibration gas by [22] .  Qmix — ~, p7~\ * Qref [P^pjmix  (E-1)  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 i n the total flow. rh  , rh  He  Pmix =  air  ~.  X  p  He  +  —  _X  (F.2)  Pair  rh = m e + rh T  H  (F.3)  air  The expression for the mixture C is similar to Equation F.2. The mass fractions correp  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) »i > - . 6 4 . 4 % air 0.8(4.003 g/mol) + 0.2(28.97 g/mol)  , , (F.5)  M  v  }  Therefore, the respective mixture properties are calculated to be Pmix = 0.356(0.1615 kg/m ) + 0.644(1.169 kg/m ) - 0.8.10 kg/m 3  {C ) ix p  m  3  3  = 0.356(5.193 kJ/kgK) + 0.644(1.004 kJ/kgK) = 2.495 kJ/kgK  (F.6)  (F.7)  Finally, the corrected actual flow rate of the dual gas mixture is (1.169 kg/m )(l.004kJ/kgK)  ,  3  =  (0.8!0 4/m^)(2.495  J,k K) g  ^  L  P  M  )  ~  2  3  2  3  L  P  M  < » R8  F . l . Flow Rate Correction  F.l.2  157  Manufacturer-Suggested Correction  Alternatively, the manufacturer suggests a slightly different correction procedure that involves more complex C and molecular structure expressions for the mixture [23]. The p  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 is p  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 N  mix  = ^  TTIT  xN  He  +^  xN  (F.ll)  air  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 and p  N are found to be  =  1.302 U/kgK  N  mix  (  )  F  = 0.356(1.04) + 0.644(1.00) = 1.014  12  (F.13)  while the corrected mixture sample flow rate is  .1.014,(1.169 kg/m )(l.004 3  Q  kJ/kgK)  l  A  ™ " Tmr (0.81o4/m3)(1.302TO) * ' ° (  )  ( 4  n  ^  T  L  P  l  M  X  )  A  r  = 4 5 1 5  r  „  w  (  F  / T  ,  1  4  ,  A  )  The significant difference between these two correction results, as well as their discrepancies 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 i g u r e 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. T a b l e F . l : Calibrated Flow Rates ( L P M ) Flow Meter Setting (LPM)  2.0 3.0 4.0 5.0  100 2.92 3.96 4.93 5.65  Helium Fraction (vol. %) 90 80 2.69 2.49 3.82 3.54 4.68 4.53 5.44 5.28  0 1.84 2.73 3.68 4.55  F.2. Experimental Calibration  160  Time (s)  Flow Rate (LPM}  100-400 100-400 100-400 100-400 100-400  T a b l e F . 2 : Calibration Data Time Flow Rate Setting Volume Range (LPM) (LPM) (mL) (s) Ambient Air 9.68 1.86 2.0 100-800 9.66 1.86 2.0 100-800 9.72. 1.85 2.0 100-800 9.69 1.86 2.0 100-800 9.63 1.87 2.0 100-800  22.16 22.03 22.12 22.25 22.06  1.90 1.91 1.90 1.89 1.90  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  6.65 6.59 6.62 6.66 6.53  2.71 2.73 2.72 2.70 2.76  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  15.19 15.07 15.12 15.09 15.09  2.76 2.79 2.78 2.78 2.78  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  5.00 4.91 5.03 4.97 4.94  3.60 3.67 3.58 3.62 3.64  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  11,31 11.40 11.41'" 11.31 11.32  3.71 3.68 3.68 3.71 3.71  5.0 5.0 .5.0 5.0 5.0  100-400 100-400 100-400 100-400 100-400  4.06 4.03 4.16 4.16 4.10  4.43 4.47 4.33 4.33 4.39  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  9.18 9.19 9.16 9.12 9.06  4.58 4.57 4.59 4.61 4.64  2.0 2.0 2.0 2.0 2.0  100-400 100-400 100-400 100-400 100-400  6.22 6.31 6.22 6.25 6.28  100% Helium 2.89 2.0 2.85 2.0 2.89 2.0 2.88 2.0 2.87 2.0  100-800 100-800 100-800 100-800 100-800  14.28 14.18 14.15 14.19 14.09  2.94 2.96 2.97 2.96 2.98  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  4.59 4.63 4.63 4.62 4.59  3.92 3.89 3.89 3.90 3.92  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  10.43 10.47 10.44 10.47 10.53  4.03 4.01 4.02 4.01 3.99  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  3.59 3.72 3.72 3.72 3.75  5.01 4.84 4.84 4.84 4.80  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  8.38 8.40 8.32 8.41 8.50  5.01 5.00 5.05 4.99 4.94  5.0 5.0 5.0 5.0 5.0  100-400 100-400 100-400 100-400 100-400  3.19 3.19 3.22 3.22 3.22  5.64 5.64 5.59 5.59 5.59  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  7.37 7.38 7.28 7.40 7.43  5.70 5.69 5.77 5.68 5.65  Setting (LPM)  Volume Range (mL)  2.0 2.0 2.0 2.0 2.0  F.2. Experimental'Calibration Setting (LPM)  Volume Range (mL)  Time  2.0 2.0 2.0 2.0 2.0  100-400 100-400 100-400 100-400 100-400  9.91 9.94 9.85 9.94 10.00  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  5.0 5.0 5.0 5.0 5.0  161 Flow Rate  Volume Range  Ambient Air 1.82 2.0 1.81 2.0 1.83 2.0 1.81 2.0 1.80 2.0  Time (s)  100-800 100-800 100-800 100-800 100-800  22.75 22.72 22.78 22.66 22.60  1.85 1.85 1.84 1.85 1.86  6.62 6.65 6.71 6.59 6.60  2.72 2.71 2.68 2.73 2.73  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  15.09 15.16 15.13 15.10 15.13  2.78 2.77 2.78 2.78 2.78  4.94 4.94 4.91 4.90 4.87  3.64 3.64 3.67 3.67 3.70  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  11.31 11.22 11.25 11.19 11.19  3.71 3.74 3.73 3.75 3.75  100-400 100-400 100-400 100-400 100-400  3.94 3.97 3.90 3.85 4.06  4.57 4.53 4.62 4.68 4.43  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  9.10 9.10 ; 9.10 9.07 9.06  4.62 4.62 4.62 4.63 4.64  2.0 2.0 2.0 2.0 2.0  100-400 100-400 100-400 100-400 100-400  6.81 6.75 6.71 6.81 6.78  90% Helium 2.64 2.0 2.67 2.0 2.68 2.0 2.64 2.0 2.65 2.0 -  100-800 100-800 100-800 100-800 100-800  15.44 15.41 15.41 15.50 15.40  2.72 2.73 2.73 2.71 2.73  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  4.68 4.78 4.71 4.78 . 4.85  3.85 3.77 3.82 3.77 3.71  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  10.87 10.94 10.88 10.94 10.87  3.86 3.84 3.86 3.84 3.86  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  3.85 3.81 3.85 3.90 3.88  4.68 4.72 4.68 4.62 4.64  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  8.97 9.03 8.88 8.91 8.87  4.68 4.65 4.73 4.71 4.74  5.0 5.0 5.0 5.0 5.0  100-400 100-400 100-400 100-400 100-400  3.32 3.35 3.32 3.34 3.28  5.42 5.37 5.42 5.39 5.49  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  7.66 7.68 7.66 7.69 7.75  5.48 5.47 5.48 5.46 5.42  s  Setting  Flow Rate| (LPM)  i  F.2. Experimental Calibration 1 Setting (LPM)  I  162  Volume Range (mL)  Time (s)  2.0 2.0 2.0 2.0 2.0  100-400 100-400 100-400 100-400 100-400  9.97 10.00 10.06 9.93 10.00  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  6.72 6.68 6.71 6.75 6.72  2.68 2.69 2.68 2.67 2.68  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  4.94 4.97 4.96 4.94 4.97  5.0 5.0 5.0 5.0 5.0  100-400 100-400 100-400 100-400 100-400  2.0 2.0 2.0 2.0 2.0  Flow Rate Setting (LPM) (LPM) Ambient Air 1.81 2.0 1.80 2.0 1.79 2.0 1.81 2.0 1.80 2.0  Volume Range (mL)  Time (s)  Flow Rate (LPM)  100-800 100-800 100-800 100-800 100-800  22.94 22.94 22.90 22.91 22.94  1.83 1.83 1.83 1.83 1.83  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  15.40 15.47 15.34 15.43 15.37  2.73 2.71 2.74 2.72 2.73  3.64 3.62 3.63 3.64 3.62  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  11.35 11.38 11.31 11.34 11.28  3.70 3.69 3.71 3.70 3.72  3.97 3.97 4.00 3.97 4.00  4.53 4.53 4.50 4.53 4.50  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  9.12 9.13 9.19 9.15 9.13  4.61 4.60 4.57 4.59 4.60  100-400 100-400 100-400 100-400 100-400  7.29 7.34 7.34 7.31 7.31  80% Helium 2.47 2.0 2.45 2.0 2.45 2.0 2.46 2.0 2.46 2.0  100-800 100-800 100-800 100-800 100-800  16.62 16.59 16.68 16.68 16.69  2.53 2.53 2.52 2.52 2.52  3.0 3.0 3.0 3.0 3.0  100-400 100-400 100-400 100-400 100-400  5.16 5.16 5.15 5.13 5.15  3.49 3.49 3.50 3.51 3.50  3.0 3.0 3.0 3.0 3.0  100-800 100-800 100-800 100-800 100-800  11.72 11.69 11.72 11.75 11.72  3.58 3.59 3.58 3.57 3.58  4.0 4.0 4.0 4.0 4.0  100-400 100-400 100-400 100-400 100-400  4.00 4.09 4.04 4.00 3.97  4.50 4.40 4.46 4.50 4.53  4.0 4.0 4.0 4.0 4.0  100-800 100-800 100-800 100-800 100-800  9.13 9.13 9.22 9.16 9.16  4.60 4.60 4.56 4.59 4.59  5.0 5.0 5.0 5.0 5.0  100-400 100-400 100-400 100-400 100-400  3.50 3.50 3.50 3.50 3.47  5.14 5.14 5.14 5.14 5.19  5.0 5.0 5.0 5.0 5.0  100-800 100-800 100-800 100-800 100-800  7.75 7.72 7.78 7.84 7.78  5.42 5.44 .5.40 5.36 5.40  F.2. Experimental Calibration  6.00  <  163  Flow Rate Correlation for 80% Helium  i  2.00 1.00 0.00  -I  ,  1.0  2.0  —  ,  ,  ,  3.0 4.0 Setting (LPM)  5.0  1  60  Flow Rate Correlation at 5.0 LPM Setting 5.70  T  5.20  -1  ,  ,  80  85  90  :  —  n  95  Helium %  F i g u r e F . 2 : Linear Flow Rate Correlations  1 100  F.2. Experimental Calibration  F i g u r e F . 3 : 2-D Surface Interpolation  164  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. T a b l e G . l : Premixed Methane Series Experimental Test Matrix  1  *  S  u .£ fi  0.2 0.3 0.5 0.7 1 1.3 1  1050  1150  Temperature (K) 1200 1250  1300 X X  X X  7  2  1100  X  XXXXXX  X  X  Note: each X represents one experiment; all experiments at 30 bar pressure  165  1350 XXX  G.3. Error Analysis  166  1 0.5 0 -0.5  4  (  —r-  0.5  10  12  14  16  10  12  14  16  1  1  ra o >  a5-0.5 g 0  1  1  I  i  i  i  0.5  o •—  i  -0.5  —  1  —  I........  •  —  :  . . .  i—  .  1  i  i  1  i  16  10 5 0 -5 6  8 Time (ms)  10  12  14  F i g u r e 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 measurement errors in driven gas pressures. O n 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.  16  G.3. Error Analysis  167  y= 1.0351x- 0.2743 R =1 2  4.5  F i g u r e 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 introduced 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  Pmethane ^measured,2 ^measured,! ± 0.004  (G.l)  (G.2)  G.3. Error Analysis  168  T a b l e G . 2 : Premixed Methane/Ethane Series Experimental Test Matrix  0.5  .2  13 cn o)  1  ' 2 3 4  1 ro >  •  '§• LU  Temperature (K) 1100 X  1000 X  XXXX XX XX X  .  1200  xooo +  5  6  X  Note: X, 0, and + represent 40, 16, and 30 bar experiments, respectively  T a b l e G . 3 : Premixed Blank Series Experimental Test Matrix  -o  1100 XX  17  Temperature (K) 1200  xxxxx xxxxx  <D  5?  (/) <u *  30 XXX XX  40  Note: each X represents one experiment  Pair  fmea3ured,3  Pmeasured,2 ± 0.004  Ptotal ~ Pmeasuredfi ±0.002  (G.3)  (G.4)  The errors in the individual mole fractions of the driven gases are therefore:  Vethane  Pethane  Pmeasured,\ ± 0.002  Ptotal  Pmeasured,3 ±0.002  _ ^methane _ Pmeasured,2 ~~ Pmeasured.l.^ 0.004 ymethane ;— ~~ ~ — total *measured 3 ±0.002 _ Pair _ Pmeasured,3 ~ Pmeasured,2 ± 0.004 Vair — ~ j • •'total ^ measured,3 ± 0.002 D  (G.5)  (G.6)  }  (G.7)  G.3. Error Analysis  169  T a b l e G . 4 : Premixed.Methane Series Conditions Number  Name  M1 M2 M3 M4 M5  pnopmOl p30s01 p30s02 p30s03 p30s04 p30s05 p30s06 p30s07 p30s08 p30s09 p.30s10 p30s11 p30s12 p30s13 p30s14 p30s15  ,M6  M7 M8 M9 M10 M11 M12 M13 M14 M15 M16  Methane Series Methane Equivalence Mass (g) Ratio n/a n/a 0.433 1.02 0.302 0.75 0.212 0.53 0.583 1.40 0.125 0.32 0.773 1.81 0.915 2.10 0.082 0.21 0.082 0.21 0.082 0.21 0.434 1.05 0.438 1.06 0.421 1.00 0.422 1.00 0.422 1.00  Pressure (bar) 29.7 27.7 27.3 31.3 27.7 28.7 29.2 29.6 24.3 22.6 22.5 22.0 23.9 23.7 23.3 24.3  Temperature (K) 1194 1129 1186 1311 1100 1308 1074 1045 1214 1173 1169 1026 1067 1060 1050 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:  Methane EQR  2(j^methane)  PmeasuredA ± 0.002 ±0.004 P easured,2 ~ Pmeasured,!  0.21(t/ j ) a r  Methane/Ethane  (G.8)  m  EQR = ^(Vmethane) + ^-^{yethane) 0.21(y ) air  ^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 errors 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. T h e total fuel error is 0.0112 grams (1.11%). T h e resultant error in E Q R is calculated to be 0.023.  G.3. Error Analysis  G.3.2  170  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. E = 0.002(d d  + 0.002  )  mea3ured  £  '=i24(i=  8 x 1 0  "  (G.10)  6  ( G  - > n  The combined error in the incident shock velocity (V) is therefore  dV E  total  =  ^ ( — y { E  E tai = yj(^) (0.002(d  )  2  to  measured  dV d  y  +  (G.12)  {-^y(E y t  + 0.002) + ( ^ ) ( 8 x l O " ) 2  2  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 cumulative 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 c m , its diameter is 2  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  Qsample ~ Qmeasured ± 0 ' 0 2 ( Q  (G.14)  measurec  ()  (G.15)  Dtapespot = 7.96 mm ± 0.04 m m  (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 9.39.6% 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  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  Methane Mass(g) 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  Methane/Ethane Series Ethane Methane Equivalence Mass (g) Fraction (mol %) Ratio n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a 0.080 91.7% 1.009 0.102 91.2% 1.008 0.132 90.9% 1.010 0.118 90.0% 1.034 0.221 88.5% 2.099 0.243 89.9% 3.163 0.294 89.4% 4.270 0.352 89.2% 6.723 0.118 90.2% 1.053 0.066 90.0% 0.513 0.110 90.7% 1.037 0.211 88.8% 2.085 0.239 90.2% 3.238  Note: n/a indicates insufficient information was recorded for the calculation  Pressure (bar) 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  Temperature! 1125 1270 1182 1239 1174 1117 1018 1024 1044 1034 1089 1031 1079 1072 1047 1061 1107  G.3. Error Analysis  173  T a b l e G . 6 : Premixed Blank Series Conditions >  c c  TO  a> u x:  tn  c 0) E E o O  03 3  TO TO  TO TO T3 X J  tt) >  O O  T3  10  C TO >  CO  IW  5  P  <D  TOTOTOTOTOTOo  o c  •<LU  T5 C T3 t ) T J b TO X  13  a>  TO  TO  r-  t  ( O CO O  B (<)  cu  s  cu ^  0) a)  "S 8% 3  TO TO  cu t:  cu (0  in  </)  3  00  a>  io o to N s o3 * nO Tm- O T - o o o  TO  c  T3  m 3  cu  O . Q . Q . O . Q . Q _  •g "g  0 O co LU T LU  o  "D  x: x: x: x: x: x:  O O  o c  o c <D  0 0 ) 0 ) 0 ) 0 ) 0 ) TO TO TO TO TO TO i_ L _ u . L . k_ i _  0) d)  C  to to to in in in E E ' E E E E  5 2 ?  CO  o  a> o <o co i n m  Q.  •- o to -= TO .55 ^ c C 0)  O 3 ~ CT TO CU  E  CU  E  w  •- cu a> co m a> o co •<-*r h-' o CD oo od co Tiro CO  cu  E  TO  Z  0) XI  E 3  Z  csi  iri iri iri  r~-'  T-CNtNCMCMCNCNCN  c  .9>  E o  ,9 t  £ fi  ° 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 )  ooSnno  w  <  /  )  <  '  >  < > /  <  '>ioto<oioto  c o. cQo .c Q o c. oQc o. c o c o c o c o E E E E E E Q . Qc .o Q a. a a a  — TO cS J2 l§  xs E  | 8CN c  o  z  a)  Methane Series  Number  Name  Aethalometer File  Date (m/d/y)  Sample Duration  Initial ATN  M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 M13 M14 M15 M16  pnopmOl p30s01 p30s02 p30s03 p30s04 p30s05 p30s06 p30s07 p30s08 p30s09 p30s10 p30s11 p30s12 p30s13 p30s14 p30s15  BC160903 BC291003 BC301003 BC301003 BC311003 BC051103 BC061103 BC071103 BC171103 BC181103 BC191103 BC201103 BC211103 BC251103 BC261103 BC271103  9/16/03 10/29/03 10/30/03 10/30/03 10/31/03 11/5/03 11/6/03 11/7/03 11/17/03 11/18/03 11/19/03 11/20/03 11/21/03 11/25/03 11/26/03 11/27/03  17:44-18:51 16:47-17:35 13:11-14:12 17:04-17:41 13:10-14:35 13:27-14:10 14:38-15:21 15:07-15:39 16:55-17:32 13:26-14:07 14:13-14:54 13:44-14:26 13:20-14:02 15:58-16:38 16:16-17:02 14:38-15:20  4.62 3.345 3.921 6.978 7.015 13.495 "7.139 1.925 13.51 12.392 1.805 3.412 19.186 18.404 10.736 0.226  Note: a = 16.6 m /g; collection spot area = 0.5 cm 2  Total Volume (L) 351.6 343.7 346.1 353.1 348.7 351.5 353.0 349.2 356.6 357.7 365.0 362.6 360.2 352.1 345.9 341.1  w  cr ST  Flow Sample BC Mass Total BC Normalized O Rate Time Increment Mass BC Mass (LPM) (min) (ng/min) (uq/q) (ng) T I 4.76 73.8 0.607 45 n/a ra 4.76 72.2 393.119 3 28370 65.465 x 4.78 72.4 414.613 30032 99.331 ra 4.85 &72.8 286.039 20826 98.218 4.82 72.3 195.663 14149 24.275 4.86 72.3 114.758 8293 66.204 cr 4.81 73.4 135.583 9951 12.865 S 4.79 72.8 179.712 13087 ra 14.299 cc 4.87 73.2 20.790 ra 1522 18.551 —i ra" 4.87 73.4 33.082 2430 29.619 cn 4.87 74.9 131.068 9820 119.777 > 4.77 76.0 88.603 6733 15.515 ra 4.76 75.6 69.782 5277 12.055 cr 4.77 73.8 54.502 4023 9.552 o" 4.77 72.5 47.629 3455 8.186 B 4.77 71.5 42.297 3023 7.161 a ra i-(  2  O o 0  Number  Name  Aethalometer File  Date (m/d/y)  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  Methane/Ethane Series Sample Initial Total Flow Duration ATN Volume Rate (LPM) (D 12:12 •13:20 31.617 346.6 4.79 15:48 16:50 52.467 227.5 4.79' 12:29-13:29 15.347 227.5 4.79 14:42- 15:35 26.615 220.9 4.62 11:34- 13:15 10.56 451.2 4.79 14:06- 15:48 119.786 452.3 4.78 17:04- 18:30 1.002 417.0 4.69 14:14- 15:38 9.106 469.3 4.72 17:16- 18:23 6.712 468.1 4.82 14:32- 15:25 29.819 468.5 4.85 17:09- 17:59 71.903 467.9 4.88 14:14- 14:30 18.023 461.2 4.87 13:16- 14:16 16.696 476.3 4.77 16:03- 17:01 36.881 473.8 4.73 11:41 12:29 2.356 473.0 4.72 13:29 14:36 1.802 471.7 4.81 15:16 15:48 0.973 470.8 4.85  Note: a = 16.6 m /g; collection spot area = 0.5 cm 2  2  Sample Time (min) 72.3 47.5 47.5 47.8 94.2 94.7 88.8 99.3 97.2 96.5 95.8 94.6 99.9 100.2 100.2 98.0 97.1  BC Mass Total BC Normalized Increment Mass BC Mass (ng/min) (ng) (ua/a^ V '3/ \ y a; 5.445 394 n/a 4.414 210 n/a 1.882 89 n/a 2.079 99 n/a 5.357 504 0.913 6.556 621 0.929 10.647 946 1.140 139.563 13863 20.316 23.149 2250 2.005 22.735 2195 1.568 37.803 3623 2.237 1155.717 109360 57.487 1439.000 143763 206.790 132.607 13281 34.609 144.098 14438 21.049 249.109 24421 22.030 4404.331 427574 302.909 u  Number  Name  Aethalometer File  Date (m/d/y)  Sample Duration  B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11 B12 B13 B14 B15 B16* B17  pnopmlO mepm07 mepm08 mepm09 mepmlO mepm11 mepm13 p30sb01 p30sb02 p30sb03 p30sb04 p30sb05 p30sb06 p30sb07 p30sb08 p30sb09 p30sb10  BC290903 BC091003 BC101003 BC141003 BC151003 BC161003 BC181003 BC231003 BC291003 BC131103 BC141103 BC191103 BC041203 BC041203 BC121203 BC171203 BC181203  9/29/03 10/9/03 10/10/03 10/14/03 10/15/03 10/16/03 10/18/03 10/23/03 10/29/03 11/13/03 11/14/03 11/19/03 12/4/03 12/4/03 12/12/03 12/17/03 12/18/03  16:46-17:35 14:55-16:18 14:14-14:41 17:21-18:13 16:13-16:45 16:40-17:30 14:11-15:10 12:32-13:36 13:26-14:06 12:43-13:30 12:37-13:26 17:02-17:52 14:17-15:02 17:04-17:40 16:49-17:30 16:55-17:35 12:21-12:59  Note: a = 16.6 m /g; collection spot area = 0.5 cm 2  2  Blank Series Initial Total ATN Volume 10.77 2.781 14.263 73.919 10.332 80.009 9.968 86.682 9.152 20.969 18.759 124.69 9.159 62.146 6.301 5.591 11.626  251.7 472.4 463.1 460.6 256.6 468.1 468.1 441.1 345.2 343.5 345.3 346.3 344.9 344.9 346.2 338.6 341.2  Flow Rate (LPM) 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  Sample Time (min) 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  BC Mass Total BC Normalized | Increment Mass BC Mass (ng/min) (ng) (ng) 132.712 7222 10044 333.781 32443 24035 207.253 20418 15431 60.359 5919 4498 35.536 1971 2688 26.358 2623 1961 15.658 1558 1165 142.730 13459 10680 121.272 8879 9002 65.881 4796 4886 23.519 1722 1746 32.538 2361 2386 41.589 3039 3084 56.801 4155 4216 27.766 2037 2059 78.918 5668 5860 50.028 3617 3711  Appendix H Non-Premixed Experiment Data H.l  Procedure  Figure H . l shows the initial J-43 injector (with 1.1 m m 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 m e t h a n e / D M E series, 9.4-9.5% for  177  H.3. Error Analysis  tn  E  f CO  tn D-  0.5 1 1.5  178  Measured Final Mass in Test Chamber Injection Pressure (psig) 834 1014 1234 1.790482666 2.046109308 2.338567234 3.800053003 4.384676542 5.407767735 5.700079504 6.138547159 8.330885429  J-43 Mass Flow Correlation  Pulse Width (ms) F i g u r e 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  T a b l e H . l : Non-Premixed M e t h a n e / D M E Series Experimental Test Matrix 1100  1200  7 CD  10  X OO  CD k_ 3  in m  20  X  Q-  25  X  £  Temperature (K) 1300 1400 O  1500  OO  XX 0 00000  1600 XXX* OOOO OOOOO  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  T a b l e H . 2 : Non-Premixed Methane Series Experimental Test Matrix  Pressure = 30 bar  1100  .  1150  xxxxx xxxx  1200 XX  Temperature (K) 1250 1300 XXXXX  1350  xxxxx XXX  000  OOOOO OOOOO OOOOO  Note: 1) X represents injection experiment; 0 represents blank experiment 2) Injection Duration = 1.5 ms 3) Injection Pressure = 75 bar  1400 XXX  H.3. Error Analysis  180  T a b l e H . 3 : Non-Premixed Methane/Ethane Series Experimental Test Matrix  Pressure = 30 bar  1100  1150  OO O  1200 XX XX  Temperature (K) 1250 1300 XX XX XX X  1350 XX XX  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  1400 XX XX  Number NMD1 NMD2 NMD3 NMD4B NMD5B NMD6B NMD7B NM08B NMD9B NMD10B NMD11B NMD12B NMD13 NMD14B NMD15 NMD16 NMD17B  Name dmeinj03 dmeinj04 dmeinj05 dmeinj07 dmeinj08 dmeinj09 dmeinjlu dmeinjH dmeinj12 dmeinj13 dmeinj14 dmeinj15 dmeinj16 dmeinj17 dmeinj18 dmeinj19 dmeinj20  Fuel Mass (mg) 3.098 3.098 3.098 0 0 0 0 0 0 0 0 0 3.098 0 1.549 3.098 0  Pressure (bar) 21.2 21.2 17.6 21.4 24.0 22.9 22.5 22.1 22.7 22.7 10.7 4.4 7.5 7.4 7.7 7.5 7.6  Non-Premixed Methane/DME  Temperatu (K) 1177 1176 1074 1184 1489 1226 1213 1202 1219 1219 1455 1255 1671 1664 1701 1635 1675  Series  Comments  Apparatus start of bag sampling  Cleaning  removed rubber gaskets dry cloth wet cloth pressurized air (plastic tubing) pressurized air (plastic tubing) brush, Simple Green, water brush, Simple Green soak, water  brush  inert driven gas (N ) 2  brush brush brush brush  NMD18B dmeinj21 0 7.7 1692 inert driven gas (N ) 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 pressurized air (multiple jet plastic tubing) NMD24B dmeinj35 0 10.5 1648 H burnout, positive pressure cleaning NMD25B dmeinj36 0 10:5 1648 filtered driven gas NMD26B dmeinj37 0 10.7 1665 NMD27B dmeinj41 0 10.2 1413 NMD28B dmeinj42 0 10.8 . 1655 filtered pressurized air NMD29B dmeinj43 0 11.1 1676 minimize vacuum pump oil condensation Note: 1) suffix B indicates a blank experiment 2) target injection durations of 1 ms (0.5ms for NMD15) will be used since the injection signal was not recorded 3) approximate injection pressures of 50 bar will be used 2  2  Diaphragm Material 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  NM1 NM2 NM3 NM4 NM5 NM6 NM7B NM8B NM9B NM10 NM11 NM12 NM13 NM14 NM15 NM16 NM17  p30nt01 p30nt04 p30nt05 p30nt06 p30nt07 p30nt08 p30nt09b p30nt10b p30nt11b p30nt12 p30nt13 p30nt15 p30nt16 p30nt17 p30nt18 p30nt19 p30nt20  Fuel Mass (mg) 7.666 8.224 8.480 8.368 7.624 7.964 0 0 0 7.773 7.996 7.629 7.519 8.033 7.352 7.624 8.331  NM18 NM19 NM20 NM21B NM22B NM23B NM24B NM25B NM26 NM27 NM28 NM29 NM30 NM31 NM32 NM33  p30nt21 p30nt22 p30nt23 p30nt24b p30nt26b p30nt27b p30nt28b p30nt29b p30nt30 p30nt31 p30nt32 p30nt33 p30nt34 p30nt35 p30nt36 p30nt37  7.959 7.773 7.885 0 0 0 0 0 7.662 8.108 7.662 8.220 8.034 7.550 7.550 7.671  Non-Premixed Methane Series Pressure Temperature Injection Apparatus Comments (bar) (K) Pressure (bar) 30.6 1422 75 10-minute bag settling before sampling 30.0 1300 75 30.6 1210 . 75 29.2 1133 75 29.3 1293 75 28.9 1152 75 29.0 1178 n/a 28.5 1168 n/a 28.5 1168 n/a 28.3 1262 75 28.5 1265 75 29.8 1293 74 29.7 1293 74 29.7 1293 74 30.3 1305 72 30.4 1307 75 30.1 1300 75 29.1 1280 75 29.7 1293 75 30.1 1300 75 29.9 1300 n/a 30.2 1305 n/a start of 1/2" single-stage impactor 30.0 1298 n/a 30.1 1300 n/a 29.8 1293 n/a 29.7 1142 75 30.5 1158 75 30.4 1157 75 29.8 1147 75 29.6 1143 75 30.1 1153 75 28.2 1353 75 30.2 1402 75  cr  ST Or  "Z a o  — it  a>  a x' CD  D-  o  c-t-  0  CB'  CO  Con  Name  e Ser  Number  a. c-t-'  0' CO  Number  Name  NM34 NM35 NM36B NM37B NM38B NM39B NM40B NM41B NM42B NM43B NM44B NM45B  p30nt38 p30nt39 p30nt40b p30nt41b p30nt42b p30nt43b p30nt44b p30nt45b p30nt46b p30nt47b p30nt48b p30nt49b  Fuel Mass (mg) 7.848 7.997 0 0 0 0 0 0 0 0 0 0  Pressure (bar) 30.7 30.2 29.7 30.1 29.8 30.0 30.3 30.4 30.3 28.5 30.6 30.7 •  Temperature (K) 1216 1152 1293 1300 1293 1300 1305 1307 1307 1263 1312 1312  Note: 1) suffix B indicates a blank experiment 2) 0.030" Lexan diaphragms used for all experiments 3) n/a = not applicable  Injection Pressure (psi) 75 75 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a  Apparatus Comments  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 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  9°  Apparatus/Procedure Comments  >-! —I  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  o  V cr ST  > 3 co co'  •z 3  i  T)  3  P ra c-t-  tr 3  a>  Ul CD  >-!  co'  co  o  O 3 QO 3 co  OO 4^  Non-Premixed Methane/DME Series  Number  Name  Aethalometer File  Date (m/d/y)  Sample Duration  Initial ATN  NMD1 NMD2 NMD3 NMD4B NMD5B NMD6B NMD7B NMD8B NMD9B NMD10B NMD11B NMD12B NMD13 NMD14B NMD15 NMD16  dmeinj03 dmeinj04 dmeinj05 dmeinj07 dmeinj08 dmeinj09 dmeinjIO dmeinjH dmeinj12 dmeinj13 dmeinj14 dmeinj15 dmeinj16 dmeinj17 dmeinj18 dmeinj19  BC100703 BC110703 BC110703 BC140703 BC160703 BC160703 BC160703 BC170703 BC170703 BC 180703 BC180703 BC180703 BC210703 BC210703 BC210703 BC210703  7/10/03 7/11/03 7/11/03 7/14/03 7/16/03 7/16/03 7/16/03 7/17/03 7/17/03 7/18/03 7/18/03 7/18/03 7/21/03 7/21/03 7/21/03 7/21/03  15:36-16:17 13:10-13:50 16:21-16:36 13:17-13:53 10:31-11:17 12:48-13:18 15:24-16:01 11:23-12:09 16:05-16:49 15:13-15:47 15:58-16:17 17:02-17:14 11:21-11:40 12:58-13:16 14:52-15:10 16:33-16:57  1.893 7.432 19.222 0.148 4.692 3.837 13.473 0.928 9.175 9.953 13.968 18.58 1.995 4.67 7.861 11.363  NMD17B NMD18B NMD19 NMD20 NMD21B NMD22B NMD23B NMD24B NMD25B NMD26B NMD27B NMD28B NMD29B  dmeinj20 dmeinj21 dmeinj23 dmeinj29 dmeinj30 dmeinj31 dmeinj32 dmeinj35 dmeinj36 dmeinj37 dmeinj41 dmeinj42 dmeinj43  BC220703 BC220703 BC240703 BC250703 BC250703 BC250703 BC280703 BC300703 BC300703 BC300703 BC310703 BC010803 BC010803  7/22/03 .7/22/03 7/24/03 7/25/03 7/25/03 7/25/03 7/28/03 7/30/03 7/30/03 7/30/03 7/31/03 8/1/03 8/1/03  10:51-11:08 12:00-12:21 12:04-12:46 11:15-11:36 13:25-13:45 17:08-17:32 12:29-12:50 11:17-11:42 15:26-15:44 16:36-16:55 16:04-16:26 15:12-15:33 17:08-17:27  0.068 2.297 0.497 0.854 9.922 20.283 2.167 1.088 10.565 16.18 35.596 15.027 27.943  Note: 1) a = 16.6 m /g; collection spot area = 0.5 cm 2) n/a = not applicable 2  2  Total Volume (L) 239.7 252.2 246.8 255.8 273.6 254.0 277.2 268.3 264.7 270.1 118.1 69.9 93.1 93.1 91.3 94.9 93.1 93.1 . 273.6 114.5 116.3 123.5 123.5 123.5 127.1 ' 123.5 112.8 121.7 116.3  Flow Rate (LPM) 4.70 4.70 4.57 • 4.70 4.83 4.78 4.78 4.78 4.70 4.70 4.76 4.72 4.84 4.84 4.84 4.75 4.92 4.84 4.60 4.78 4.74 4.78 4.78 4.95 4.95 4.95 4.91 4.95 4.95  Sample Time (min) 51.0 53.6 54.0 54.4 56.7 53.1 58.0 56.1 56.3 57.4 24.8 14.8 19.2 19.2 18.9 20.0  BC Mass Increment (ng/min) 3.411 ' 4.032 3.131 5.486 6.765 1.791 10.947 3.860 3.881 3.452 7.042 9.082 2.621 3.006 2.153 3.467  18.9 19.2 59.5 23.9 24.6 24.1 24.1 23.3 24.0 23.3 23.0 23.0 21.9  3.237 2.155 3.9793 13.7536 12.629 14.845 10.3347 5.6996 7.1378 7.6321 11.7984 18.193 4.788  Total BC Normalized Mass BC Mass (uq/q) (ng) \"a*a/ 174 56.136 216 69.829 169 54.604 298 n/a 383 n/a 95 n/a 635 n/a 217 n/a 219 n/a 198 n/a 175 n/a 134 n/a 50 16.283 58 n/a 41 26.231 69 22.333 61 n/a 41 n/a 237 76.478 329 106.326 310 n/a 358 n/a 249 n/a 133 n/a 171 n/a 178 n/a 271 n/a 418 n/a 105 n/a  Number  NM1 NM2 NM3 NM4 NM5 NM6 NM7B NM8B NM9B NM10 NM11 NM12 NM13 NM14 NM15 NM16  Name  . p30nt01 p30nt04 p30nt05 p30nt06 p30nt07 p30nt08 p30nt09b p30nt10b p30nt11b p30nt12 p30nt13 p30nt15 p30nt16 p30nt17 p30nt18 p30nt19  N o n - P r e m i x e d Methane S e r i e s  Aethalometer File  Date (m/d/y)  Sample Duration  Initial ATN  BC030304 BC050304 BC080304 BC080304 BC100304 BC100304 BC110304 BC110304 BC160304 BC160304 BC170304 BC180304 BC180304 BC190304 BC230304 BC230304  3/3/04 3/5/04 3/8/04 3/8/04 3/10/04 3/10/04 3/11/04 3/11/04 3/16/04 3/16/04 3/17/04 3/18/04 3/18/04 3/19/04 3/23/04 3/23/04  17:03-17:20 10:48-11:25 11:41-12:23 16:48-17:24 11:36-12:07 15:52-16:27 13:26-13:56 14:22-15:00 11:13-11:51 14:39-15:16 10:15-10:51 10:35-11:15 12:05-12:43 10:56-11:31 11:38-12:01 15:37-16:16  3/23/04 -16:51-17:29 3/24/04 11:10-11:46 3/24/04 16:00-16:33 3/24/04 16:52-17:31 3/25/04 13:15-13:53 3/25/04 16:22-16:58 3/29/04 10:08-10:39 3/29/04 10:58-11:22 3/29/04 11:39-12:00 3/31/04 11:13-11:46 3/31/04 16:38-17:09 3/31/04 17:37-18:09 4/6/04 16:45-17:00 4/7/04 11:34-11:44 4/7/04 16:26-16:55 4/8/04 13:24-13:47  NM17 p30nt20 BC230304 NM18 p30nt21 BC240304 NM19 p30nt22 BC240304 NM20 p30nt23 BC240304 NM21B p30nt24b BC250304 NM22B p30nt26b BC250304 NM23B p30nt27b BC290304 NM24B p30nt28b BC290304 NM25B p30nt29b" BC290304 NM26 p30nt30 BC310304 NM27 p30nt31 BC310304 NM28 p30nt32 BC310304 NM29 p30nt33 BC060404 NM30 p30nt34 BC070404 NM31 p30nt35 BC070404 NM32 j30nt36 BC080404  45.371 0.209 12.758 3.326 8.934 2.206 7.672 0.881 6.822 12.235 9.937 1.131 0.687 7.392 11.124 14.553  Total Volume (L) 336.4 345.7 333.3 332.9 338.0 332.7 348.8 348.0 348.7 348.9 350.3 355.7 357.2 350.9 353.4 353.4  Flow Rate (LPM) 4.79 4.81 4.77 4.71 4.81 4.73 4.63 4.71 4.71 4.76 4.76 4.81 4.81 4.81 4.81 4.81  1.838 2.811 5.615 0.139 2.303 10.916 6.404 15.11 0.27 10.814 3.413 7.008 10.37 11.33 30.146 1.754  353.4 354.7 354.0 354.7 354,7 354.7 351.9 352.7 351.9 349.3 350.1 350.0 351.0 349.3 348.5 352.6  4.81 4.81 4.81 4.81 4.72 4.81 4.73 4.81 4.89 4.73 4.73 4.65 4.73 4.73 4.65 4.81  Sample Time (min) 70.2 71.8 69.9 70.7 70.2 70.4 75.4 73.8 74.0 73.2 73.6 74.0 74.3 73.0 73.5 73.5 73.5 73.8 73.6 73.8 75.1 73.8 74.4 73.4 72.0 73.8 74.0 75.3 74.3 73.8 74.9 73.3  BC Mass Total BC Normalized Increment Mass BC Mass va / (ng/min) (ua/a) (ng) 18.464 169.086 1296 117.558 8441 1026.316 19.766 1381 162.887 13.251 937 111.948 17.247 1211 158.835 53.509 3766 472.928 13.021 981 n/a 14.219 1050 n/a 14.240 1053 n/a 12.405 908 116.868 16.566 1219 152.478 12.778 946 123.953 16.453 1223 162.599 15.658 1143 142.333 13.051 960 130.517 35.665 2622 343.916 28.806 2118 254.215 21.111 1558 195.754 17.524 1291 166.033 24.283 1792 227.292 27.576 2071 n/a 9.316 688 n/a 6.009 447 n/a 4.785 351 n/a 6.914 498 n/a 4.083 301 39.342 3.418 253 31.189 2.705 204 26.570 2.593 192 23.420 5.443 402 50.018 6.431 482 63.838 5.512 404 53.540  Number  IN M O O  NM34 NM35 NM36B NM37B NM38B NM39B NM40B NM41B NM42B NM43B NM44B NM45B  Name  p30nt37 p30nt38 p30nt39 p30nt40b p30nt41b p30nt42b p30nt43b p30nt44b p30nt45b p30nt46b p30nt47b p30nt48b p30nt49b  Aethalometer File  Date (m/d/y)  BC080404 BC080404 BC080404 BC 120504 BC130504 BC140504 BC140504 BC170504 BC 180504 BC 180504 BC190504 BC200504 BC200504  4/8/04 4/8/04 4/8/04 5/12/04 5/13/04 5/14/04 5/14/04 5/17/04 5/18/04 5/18/04 5/19/04 5/20/04 5/20/04  Note: 1) a = 16.6 m /g; collection spot area 2) n/a = not applicable 2  :  o  Non-Prerrnxed Methane Series (Continued)  Sample Duration  Initial ATN  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  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  0.5 cm  2  Total Volume (L) 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  Flow Rate (LPM), 4.85 4.77 4.73 4.80 4.80 4.80 4.80 4.80 4.80 4.80 4.88 4.80 4.80  Sample Time (min' (min) 72.8 73.7 74.0 73.5 73.3 73.1 73.1 73.2 73.3  73.2  72.0 73.1 73.1  i  BC Mass Total BC Increment Mass (ng/min) 'min) (ng) 5.946 433 4.813 355 7.021 519 9.896 727 22.510 1651 19.137 1399 48.875 3573 13.926 1019 10.320 757 22.156 1621 16.104 1159 22.287 1629 14.287 1044  Normalized BC Mass  -d CD  X  56.454 45.181 64.951 n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a  CD  ra ra  > ra cr  SL o* B  ra c-t-  CD •-!  o  o  o.  Number  Name  Aethalometer File;  Date (m/d/y)  NME1B NME2B NME3B NME4B NME5B NME6B NME7B NME8B NME9B NME10 NME11 NME12 NME13 NME14 NME15 NME16  p30nts04b p30nts05b p30nts13b p30nts19b p30nts22b p30nts23b p30nts26b p30nts30b p30nts35b p3ntsb01 p3ntsb03 p3ntsb04 p3ntsb05 p3ntsb06 p3ntsb07 p3ntsb08  BC110604 BC110604 BC220604 BC240604 BC250604 BC250604 BC280604 BC290604 BC300604 BC180804 BC190804 BC190804 BC190804 BC230804 BC230804 BC230804  NME17 NME18 NME19 NME20 NME21 NME22 NME23 NME24 NME25 NME26 NME27 NME28  p3ntsb09 p3ntsb10 p3ntsb11 p3ntsb12 p3ntsb13 p3ntsb14 p3ntsb15 p3ntsb16 p3ntsb17 p3ntsb18 p3ntsb19 p3ntsb20  BC240804 BC240804 BC240804 BC250804 BC250804 BC250804 BC260804 BC260804 BC260804 BC270804 BC270804 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 m /g; collection spot area = 0.5 cm 2  Non-Premixed Methane/EthaneSeries Sample Initial Total Flow Duration ATN Volume Rate LPM) 12:00-12:43 11.834 350.8 4.85 14:03-14:55 9.422 350.1 4.85 13:21-14:13 3.133 351.5 4.85 13:03-13:49 6.011 350.1 4.84 12:05-12:57 4.722 347.9 4.85 13:28-14:03 348.7 0.889 4.85 13:04-14:00 348.4 1.593 4.77 12:17-12:58 348.8 0.127 4.85 12:24-13:04 347.3 4.544 4.78 14:48-15:30 348.6 19.378 4.85 10:46-11:54 348.7 0.933 4.84 14:18-15:15 349.4 5.285 4.76 15:55-16:13 350.6 0.123 4.72 12:16-13:11 351.8 9.512 4.73 15:27-16:07 352.6 16.256 4.73 16:42-17:00 351.1 22.01 4.73 11:58-13:03 15.191 356.3 4.81 13:28-14:28 22.563 357.1 4.76 15:18-15:47 27.949 356.4 4.63 12:23-13:19 14.003 357.8 4.81 14:18-15:20 21.113 356.1 4.70 15:55-16:05 24.901 356.0 4.70 12:40-13:35 5.166 353.3 4.87 14:05-15:00 15.799 4.79 353.3 7.045 15:44-16:01 352.1 4.82 12:30-13:30 12.202 350.8 4.82 14:25-15:08 21.057 350.8 4.83 5.258 15:52-15:57 350.0 4.83 2  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 Total BC Increment Mass (ng/min) (nc 17.440 1260 2.601 188 4.807 348 3.188 231 2.869 206 3.148 226 2.200 161 2.429 175 2.825 205 15.011 1079 6.791 489 1.896 139 2.415 179 3.100 231 2.584 193 3.562 265 354 4.775 195 2.595 176 2.2808 303 4.0708 198 2.6092 234 3.0903 293 4.0451 205 2.7753 330 4.512 228 3.1301 172 2.3608 422 5.8132  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