UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Morphology and microstructure of diesel particulates Soewono, Arka 2008

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

Item Metadata

Download

Media
24-ubc_2008_fall_soewono_arka.pdf [ 5.2MB ]
Metadata
JSON: 24-1.0070805.json
JSON-LD: 24-1.0070805-ld.json
RDF/XML (Pretty): 24-1.0070805-rdf.xml
RDF/JSON: 24-1.0070805-rdf.json
Turtle: 24-1.0070805-turtle.txt
N-Triples: 24-1.0070805-rdf-ntriples.txt
Original Record: 24-1.0070805-source.json
Full Text
24-1.0070805-fulltext.txt
Citation
24-1.0070805.ris

Full Text

MORPHOLOGY AND MICROSTRUCTURE OF DIESEL PARTICULATES  by ARKA SOEWONO B.A.Sc., The University of British Columbia, 2006  A THESIS SUBMITTED iN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE  in  THE FACULTY OF GRADUATE STUDIES (Mechanical Engineering) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2008  ©Arka Soewono, 2008  ABSTRACT  The effect of particulate matter on the climate depends on its scattering properties (influenced by morphology) and refractive index (dependent on microstructure). The morphology and microstructure of particulates from two different engines were studied. The first set of soot samples was collected from a 1 .9L Volkswagen Turbo Direct Injection engine with or without a catalytic converter, using two different fuel types (ULSD and B20) and six speed/load combinations. The second set of the samples was taken from a Cummins ISX heavy-duty engine using the Westport pilot-ignited directinjection natural-gas fuelling system for three different speed/load combinations.  The soot morphology was investigated using transmission electron microscopy (TEM), emphasizing the fractal properties. A Matlab-based image processor was used to extract geometrical properties of soot. Thirty-five aggregates were analyzed for each sample. The fractal dimensions (Di) were computed using the aggregate dimension and pair correlation methods. For the soot collected from VW engine, it was found that fuel type (ULSE) and B20) and the presence of a catalytic converter did not affect the fractal dimensions of soot aggregates, while engine load conditions had slight influence on Df. At constant engine RPM, fractal dimensions started to decline as the engine torque was significantly increased. For the soot produced by the natural gas engine, similar changes in the fractal dimension with respect to the engine load were also observed. Both methods of computing Df produced similar trends.  11  Raman spectroscopy was used to quantify the degree of structural disorder present in soot. The Raman spectral analysis was done using two-band (“U” at -1578 and “D” l340 cm ) and five-band (U, Dl, D2, D3, D4 at about 1580,1350, 1500, 1620 and 1200 1 cm’ respectively) combinations. For the soot sampled from VW engine, the results from both methods showed that B20 soot exhibited much greater structural disorder and the degree of graphitization of the soot increased as the engine load increased. Likewise, the Raman analysis of the soot from the Cummins engine also showed that the increased in engine load condition caused increases in the degree of the structural order of soot.  111  TABLE OF CONTENTS  ABSTRACT  ii  .  TABLE OF CONTENTS  .  iv  LIST OF TABLES  vi  LIST OF FIGURES  viii  LIST OF ABBREVIATIONS ACKNOWLEDGMENTS 1.0  INTRODUCTION  x xii 1  1.1  Research Objectives  7  1.2  Thesis Organization  8  2.0  BACKGROUND INFORMATION  10  Mechanisms of Soot Formation  10  2.1 2.1.1  Soot Particle Inception and Nucleation  12  2.1.2  Particle Growth and Agglomeration  14  2.1.3  Soot Oxidation  16  2.1.4  Soot Restructuring  16  2.2  Quantifying Soot Aggregate Morphology and Microstrueture  17  2.2.1  Aggregate Size  17  2.2.2  Aggregate Shape  22  2.2.3  Mierostructure  25  3.0  DIESEL SOOT MORPHOLOGY  35  3.1  Introduction  35  3.2  Experimental Setup  36  3.2.1  Soot Samples  3.2.2  Transmission Electron Microscope  3.3  MATLAB Image Processing  36 45 46  3.3.1  Main Program  3.3.2  FunctionBinary  49 49  3.3.3  Function Calculations  51  3.3.4  Function Diary  56  3.4  Fractal Analysis  56 iv  3.4.1  Projected Aggregate Dimensions Method  57  3.4.2  Pair Correlation Method  58  Results and Discussion  3.5  59  3.5.1  Particle Size Measurement  59  3.5.2  Fractal Properties of Soot Aggregates  66  4.0  DIESEL SOOT MICROSTRUCTURE  4.1 4.2  76  Introduction  76  Experimental Section  77  4.2.1  Samples  77  4.2.2  Raman Spectroscopy  80  4.3 4.4 4.4.1 4.4.2  Raman Spectra of Soot and Graphite Raman Spectra Analysis  83 86  Two-Band Combination Method Five-band Combination Method  87 90  4.5 Results and Discussion 5.0 CONCLUSIONS AND RECOMMENDATIONS REFERENCES APPENDIX A THERMOPHORETIC SAMPLER A.2  References  94 104 109 .  126 130  APPENDIX B SAMPLE DATABASE FILES APPENDIX C MATLAB CODES FOR IMAGE PROCESSING  132  APPENDIX D MATLAB CODES FOR FRACTAL ANALYSIS  177  APPENDIX E RAMAN SPECTROSCOPY CALIBRATION  181  APPENDIX F SPECTRAL PARAMETERS OF SOOTS  184  146  v  LIST OF TABLES  Table 2.1 Real-Time Sampling Techniques  .20  Table 3.1 Volkswagen TDI Engine Specifications  36  Table 3.2 Engine Test Modes  40  Table 3.3 Properties of Test Fuels  41  Table 3.4 Cummins ISX Engine Specifications  42  Table 3.5 Test Modes for Cummins ISX Engine  42  Table 3.6 Geometric Mean Diameter for Soot from VW Engine  60  Table 3.7 Mobility Diameters and Projected Aggregate Diameters for VW Soot Table 3.8 Mean Diameters of Primary Particles for VW Soot  62  Table 3.9 Mobility Diameters and Aggregate Diameters for Cummins ISX Soot  65  Table 3.10 Mean Diameters of Primary Particles for Cummins ISX Soot Table 3.11 Df of ULSD and B20 Soot Derived from Aggregate Dimension Method  66 69  Table 3.12 D of Natural Gas Soot Derived from Aggregate Dimension Method  71  Table 3.13 Df of ULSD and B20 Soot Derived from the PCM  72  Table 3.14 D of Natural Gas Soot Derived from PCM  73  Table 3.15 Comparison of Df of Engine-Emitted Soot from TEM Measurements  75  Table 4.1 VW TDI Engine Operating Conditions for Raman Study  78  Table 4.2 ISX Cunimins Engine Operating Conditions for Raman Study  79  Table 4.3 ISX Properties of HOPG  79  63  Table 4.4 Band Combinations Tested for Curve Fitting with Two-Band Combination... 89 Table 4.5  Values for the Raman Spectra Fitted with Two-Band Combination  90  Table 4.6 Band Combinations Tested for Curve Fitting with Five-Band Combination... 92 Table 4.7 2 x Values for the Raman Spectra Fitted with Two-Band Combination Table 4.8 “D”/”G” Ratio for Soot from VW Engine  93  Table 4.9 Dl Band FWHM for Soot from VW Engine  96  95  Table 4.10 “D”I”G” Ratio and Dl Band FWHM for Soot from Cummins ISX Engine 100 Table E.1 Positions of the Contaminant Bands  182  Table E.2 Band Combinations Tested Table F. 1 Spectral Parameters for ULSD Soot (without Oxidation Catalyst)  183  Table F.2 Spectral Parameters for B20 Soot (without Oxidation Catalyst)  184 185 vi  Table F.3 Spectral Parameters for ULSD Soot (with Oxidation Catalyst)  185  Table F.4 Spectral Parameters for Natural Gas Soot from Cummins ISX Engine  186  vii  LIST OF FIGURES  Figure 1.1 Radiative Forcing of the Climate System for the Year 2000 Figure 1.2 Factors Influencing the Scattering Properties of Aerosols  4 6  Figure 2.1 Schematic of the Soot Formation Process  12  Figure 2.2 Soot Aggregates Produced by Diesel Engine Figure 2.3 Classification of Aggregate Shape  15 23  Figure 2.4 Early Model of Soot Microstructure  26  Figure 2.5 Paracrystalline Model of Soot Figure 2.6 Tetrahedral and Trigonal Hybridizations  27  Figure 3.1 Thermal Optical Analyzer Design  38  Figure 3.2 Variations of PM Composition by Engine Operating Condition Figure 3.3 Carbonaceous Fractions of Particulate Matters  39 40  Figure 3.4 OC/TC Ratios of Particulate Matters  41  Figure 3.5 Dilution System  43  Figure 3.6 Software Interface of the TEM Sample Grid Figure 3.7 Schematic of Geometrical Properties of Soot Aggregate  45  Figure 3.8 Modular Diagram of the SAIP Program  49  Figure 3.9 The Original, and Cropped and Filtered Image of the Aggregate Figure 3.10 The Binary Image, Edge Image, and Final Imposed Image Figure 3.11 The Length (Blue) and Width (Red) of the Aggregate  50  Figure 3.12 The Length (Blue) and Width (Red) of the Primary Particles Figure 3.13 The Density-Density Correlation Function of the Aggregate Figure 3.14 The Pair Correlation Function of the Aggregate  52  29  47  51 52 54 55  Figure 3.15 The Pair and Density-Density Correlation Functions of the Aggregate Figure 3.16 Illustration of the Projected Aggregate Dimension Method  56  Figure 3.17 Illustration of the Pair Correlation Method  59  58  Figure 3.18 Particle Size Distributions for Different Engine Modes 64 Figure 3.19 Fractal Analysis Based on the Aggregate Dimensions for ULSD Soot without Oxidation Catalyst 67 Figure 3.20 Fractal Analysis Based on the Aggregate Dimensions for B20 Soot without Oxidation Catalyst  67  viii  Figure 3.21 DfOfULSD and B20 Soot Obtained from Aggregate Dimension Method... 69 Figure 3.22 Df of Natural Gas Soot Obtained from Aggregate Dimension Method 71 Figure 3.23 Comparisons between Df Derived from Aggregate Dimension Method and PCM for ULSD Soot without Oxidation Catalyst 74 Figure 3.24 Comparisons between Df Derived from Aggregate Dimension Method and PCM for B20 Soot without Oxidation Catalyst  74  Figure 4.1 Calibration Raman Spectra of Silicon Figure 4.2 Contaminant Peaks in the Recorded Raman Spectra Figure 4.3 First-Order Raman Spectra of Different Carbonaceous Materials Figure 4.4 Exemplary Curve Fit for Raman Spectra Using Two-Band Combinations Figure 4.5 Exemplary Curve Fit for Raman Spectra Using Five-Band Combinations Figure 4.6 “D”/”G” Band Intensity Ratios for Soot from VW Engine Figure 4.7 Dl Band FWHM for Soot from VW Engine  81 82 84 ....  88  ....  91 97 97  Figure 4.8 “D”/”G” Band Intensity Ratios for Soot from Cummins ISX Engine Figure 4.9 Dl Band FWHM for Soot from Cummins ISX Engine  101  Figure 4.10 D3/G Band Intensity Ratios for Soot from VW Engine Figure A. 1 Teflon TPS with Brass TEM Holder  103  Figure B.1 The TEM Original Image Figure B.2 The Cropped and Filtered Image  102 129 132 133  Figure B.3 The Binary Image Figure B.4 The Edge Image  133  Figure B.5 The Final Imposed Image  134  Figure B.6 The Length (Blue) and Width (Red) Image Figure B.7 The Length (Blue) and Width (Red) of the Primary Particles Figure B.8 The Density-Density Correlation Function  135 136  Figure B.9 The Pair Correlation Function  136  Figure B.10 The Comparison between Density Correlation and Pair Correlation  137  Figure E. 1 Contaminant Peaks in the Recorded Raman Spectra Figure E.2 Raman Spectra of Gold  181  134  135  182  ix  LIST OF ABBREVIATIONS  B20  A mixture of 20% biodiesel with 80% ultra-low sulphur diesel  DA  Projected aggregate diameter  Df  Aggregate fractal dimension  dm  Mobility diameter of aggregate Mean diameter of primary particle  DMA  Differential mobility analyzer  EC  Elemental carbon  FWHM  Full width at half maximum  HACA  Hydrogen abstraction carbon addition  HOPG  Highly oriented pyrolytic graphite  ‘Dl  The intensity of Dl band  ‘1)2  The intensity of D2 band  11)3  The intensity of D3 band  ‘1)4  The intensity of D4 band The intensity of G band  IPCC  Intergovernmental panel on climate change Fractal prefactor based on (LW)1/2  L  Projected maximum length of the aggregate  1  Projected maximum length of the primary particle  N  Number of primary particles in aggregate  OC  Organic Carbon  PAH  Polycyclic aromatic hydrocarbon  PM  Particulate matters  SAIP  Semi-automatic image processor  SMPS  Scanning mobility particle sizer  TEM  Transmission electron microscopy  x  UBC  University of British Columbia  ULSD  Ultra-low sulfur diesel  W  Maximum projected aggregate width normal to L  w  Maximum projected width of the primary particle normal t  xi  ACKNOWLEDGEMENTS  I would like to thank Dr. Steven Rogak who not only served as my supervisor but also provided invaluable feedbacks, technical expertise and encouragements throughout my academic program.  I would wish to thank Lisa Graham and co-workers (Environment Canada) as well as Scott Brown for conducting the experiment and providing diesel soot samples; Derrick Home (UBC Biolmaging Facility), Garnet Martens (UBC Biolmaging Facility), Mary Mager (UBC Metals and Materials Engineering) and Dr. Li Yang (SFU Nanolmaging Facility) for assistance with the transmission electron microscope (TEM) measurement; Dorian Tsai for our collaboration in the development of the Matlab image processor; and Dr. Dan Bizotto (UBC Chemistry) for support with the Raman spectroscope. I would also like to express my sincere thank to Jeffrey Yoslim and Christopher Laforet who have helped me one time or another during the past two years.  Last but not least, I am also grateful to my parents for their constant support and encouragement over the years.  xii  1.0  INTRODUCTION  Global warming has evolved into one of the most serious environmental challenges facing our global community. Nearly all climate scientists and most of the general public attribute the climate warming to the increase of man-made greenhouse gases (0110) in the Earth’s atmosphere (IPCC, 2001). GHG emissions result from burning of fossil fuels, industrial development, farming, and deforestation.  In contrast, aerosols are usually believed to have cooling effect both directly and indirectly on climate (Hansen et al., 1997; IPCC, 2001; IPCC, 2005). The direct cooling effect is due to the fact that most aerosols (solid or liquid particles suspended in the air) tend to reflect most of the sunlight back to the space and hence, reduce the amount of solar radiation that that actually reaches the ground (NASA, 1999).  Aerosols are also believed to have indirect cooling effect on climate through their ability to influence the properties and lifetime of clouds (Ramanathan, Crutzen, Kiehi, & Rosenfeld, 2001; Haywood & Boucher, 2000). This indirect cooling effect was likely caused by the fact that the emission of man-made aerosols serve as extra seeds planted in the atmosphere for condensation of cloud droplets (Twomey, 1977; Martin, Johnson, & Spice 1994; Boucher & Lohmann, 1995). As the concentration of the aerosols increases within a cloud, more seeds are available for the water vapor to cling to. This will lead to the creation of more water droplets, albeit smaller in size. Smaller droplets are less likely 1  to collide and fall out as rain (Albrecht, 1989). Moreover, the smaller droplets will also climb higher up into the atmosphere, and create a larger and brighter (reflecting more sunlight) cloud which has longer lifetime (Andrea et al., 2004). Larger cloud that last longer in the atmosphere will scatter more of the sun’s radiation back into space and create a local cooling effect.  However, recent studies has found evidence that certain types of aerosols containing carbonaceous particles  —  —  those  also cause a warming effect on climate (Bond &  Bergstrom, 2006; Satheesh & Ramanthan, 2000). These types of aerosols are mostly products of biomass burning or incomplete combustion of burning fuel. They usually absorb part of the solar radiation and warm the surrounding atmosphere (KaufirLan & Koren, 2006). This will cause a reduction of ambient relative humidity within the atmosphere and decrease in the temperature difference between the Earth’s surface and the upper atmosphere. This combination leads to more stable atmosphere, which suppresses the likelihood for a new cloud formation and the livelihood of existing clouds (Koren et al., 2004; Cook & Highwood, 2003; Ackerman et al., 2000). Less reflection of sunlight due to fewer clouds and less absorption of solar radiation will eventually lead to warming.  Another controversial issue surrounding the assesment of aerosol impact is related to the scale used for measuring the contribution of aerosols on climate change. Currently, the extent of aerosol impact on climate change was typically measured based on the net global change. However, this approach often leads to an erroneous claim that the cooling  2  effect of aerosols may help to balance out the warming caused by the greenhouse gases (NASA, 1999; Romanou et al., 2007) so that their influence on the climate change will be negligible. Recently, Koren (2006) suggested that the approach based on the global scale can only work properly for greenhouse gases that are well mixed in the atmosphere. On the other band, this global approach may fail to notice some of the key consequences of aerosol forcing due to the fact that aerosols are usually suspended in the atmosphere for a shorter period of time over a limited region. Hence, it is argued that the impact of the aerosol on local climate change that can be problematic. For examples local heating and cooling by aerosols may change the cloud precipitation which, in turn, can lead to a decrease in the rainfall over regions where rain is needed  —  such as rainforests  —  and shift  the precipitation over regions where the rain is not needed. These additional effects could amplify the afready problematic Earth’s greenhouse effect and lead to disastrous impacts in the long run.  The Intergovernmental Panel on Climate Change (IPCC) conducts the most comprehensive reviews of climate science. The most recent estimates of radiative forcing for aerosols, greenhouse gases and solar variability are shown in Figure 1.1. The current level of scientific understanding of each source is also listed below the axis.  3  3  The Global Mean Radiative Forcing of the Climate System for the Year 2000, relative to 1750  Halocarbons 2  Aerosols  0 2 N  4 CH 2 CO Tropospheric OZOflG  Biack carbor frorri iossi ful burnq  Mineral Dust  Q. 0  Aviation.iriducecl  Solar  Contralls Cirrus 0  Stratospheric ozone  t  Landuse &bdo only  -1 13  —2 LevelotScietditic High Medium Medium Low UII(Ie[StIHliii(J  Very Low  Very Very Very Low Low Low  Ve Low  Very Very Very Low Low Low  Figure 1.1 Radiative Forcing of the Climate System for the Year 2000 (Adapted from IPCC, 2001)  As observed from Figure 1.1, the greenhouse gases produce positive radiative forcing (warming the climate) and their impacts are generally well understood due to the fact that their spectral absorption cross-sections and concentrations in the atmosphere are accurately known (Johnson, 2003). Sulphate and organic carbon aerosol from burning fuel tend to produce a negative radiative forcing (cooling the climate) by reflecting solar radiation back to space. On the other hand, black carbon produces a positive radiative forcing because it tends to absorb the solar radiation. Biomass burning usually produces a mixture of organic carbon and black carbon (soot). The radiative forcing of biomass burning aerosol is estimated to be negative due to a high organic/black carbon ratio  4  (Johnson, 2003).  The level of scientific understanding on the climatic effect of aerosols is very low because of the total emissions of the aerosols in the atmosphere are uncertain. Furthermore, the uncertainty surrounding the organic/black carbon ratios present in the aerosols as well as the lack of knowledge on the scattering and absorption cross section of the particulates (particularly for soot) results in a high degree of uncertainties in estimates of aerosol direct forcing  Previous study done by Bond and Bergstrom (2006) has shown that the radiative forcing of carbonaceous aerosols depends on their scattering properties. As illustrated in Figure 1.2, the scattering properties of the aerosols are usually governed by both the scattering cross section and the refractive index, which in turn is highly influenced by morphology (KoylU & Faeth, 1994; Sorensen, 2001; Choi et al., 1995), and microstructure (Heckman, 1964; Schnaiter et al. 2003) respectively.  5  Scattering Cross Section Engine Type  Soot Morphology Radiative Forcing of Aerosols  Operating Condition, Fuel Type  Refractive Index Soot Microstructure  Figure 1.2 Factors Influencing the Scattering Properties of Aerosols  The fact that the radiative forcing estimates for carbonaceous particles are strongly affected by their morphological properties and micro structure has led to considerable efforts being spent on characterizing both the carbonaceous particulates morphology (Dobbins and Megaridis, 1991; Roessler & Faxvog 1981; Sorensen and Feke 1996; Lee et al. 2002; Köylu & Faeth, 1992), and their microstrueture (Rosen & Novakov 1977; Dippel & Heintzenberg, 1999; Schnaiter et a!. 2003; Saito et al. 1991). For particles emitted by engines, it is suspected that the factors influencing the nature of PM in exhaust emissions will include engine type, engine loading condition, and fuel composition. Unfortunately, there is very limited knowledge about the effect of the engine operating parameter and fuel type on the particle morphology and structure because majority of past studies have either used numerically simulated aggregates or flame-generated soot,  and only few of them are focused on particulates emitted from car engines.  Recently, with mounting evidences indicating motor vehicles as one of the major contributors of particulate matter pollution combined with growing concern of their effect  6  on climate and continuing effort to reduce the impact of particle emission from transportation sector on the atmosphere, there is an urgent need to improve scientific knowledge and information on the morphology and microstructure of engine-emitted particles. In response to this, the present research seek to contribute to a better understanding of the morphological and structural information on carbonaceous particulates, specifically those produced by both light-duty and heavy duty engines, so that extent of their influence on the climate can be better estimated in the future.  1.1 Research Objectives The overall purpose of this research is to improve the scientific knowledge and information on the aggregate morphology and n*rostructure of engine-emitted particulates as they should be useful for further assessing the climatic impact of carbonaceous particulates. The specific objectives of this study are classified into three categories.  The first objective is to characterize the morphological properties of the soot aggregate, emphasizing the fractal properties. In order to achieve this objective, the transmission electron microscopy (TEM) has been employed to image the soot aggregates. The TEM images of the aggregates have been analyzed using Matlab-based image processor to measure the projected aggregate dimensions as well as the size distribution of the primary particles. The information obtained from image processing has subsequently been used to derive the fractal properties. Furthermore, this study also represents the first attempt to estimate the fractal properties of soot produced by natural gas engine at UBC.  7  The second objective is to qualitatively compare the degree of graphitization present in the different soots, as this may correlate the optical properties with the engine type, fuel and operating conditions. The distinction of different types of soot was done via Raman spectroscopy analysis.  The last objective of this research is to provide an on-line database of well-characterized soot images available to other researchers interested in further post processing. All soot TEM images taken for this research is available online through UBC Quartz PCI database, which can be accessed from UBC Biolmaging Facility website at http://www.emlab.ubc.ca!index.html  I 2 Thesis Organization This thesis has been divided into four further chapters.  Chapter 2 provides an overview on the soot formation mechanisms, as well as literature reviews on the morphology and microstructure of carbonaceous particulates.  Chapter 3 describes the transmission electron microscopy (TEM) and two different methodology of fractal analysis that have been employed to obtain the morphological properties of soot aggregates. An overview of the Matlab-based Image processor that has been developed to extract information of the soot aggregates from the TEM images are also discussed in this chapter.  8  Chapter 4 presents the Raman spectroscopy analysis that has been performed on the soot samples. This chapter also describes two different methods of Raman spectral analysis used in this study, followed by the detailed discussion on the results obtained from both methods.  Finally, Chapter 5 summarizes the principle findings of this research.  9  2.0  BACKGROUND INFORMATION  In this chapter, literatures on the morphology and microstructure of carbonaceous particulates are reviewed. This literature review is organized into two main sections. The first section covers an overview of soot formation mechanisms, including the soot produced by engine. The second section reviews the theory on soot aggregate size, shape  and microstructure, as well as outlines several commonly used diagnostic tools for characterizing soot morphology and mierostructure.  2.1 Mechanisms of Soot Formation In general term, particulate matters (PM) are defined as any solid or liquid particles (except water) entrained in the engine exhaust gas. The largest fraction (40%-80% by mass) of the engine emitted particulates is made up of soot (Prakash, 1998); other components of diesel particulate matters include the sulfate particulates (consist mainly of hydrated sulfate acid), volatile organic fractions (mainly comprised of unburned fuel and lubricating oil) and metal abrasion particles (Kittelson, 1998).  Technically, soot is defined as black solid product of incomplete combustion or pyrolysis of hydrocarbons (fossil fuels) and other organic materials and often considered as carbonaceous particles that consists of at least 1% of hydrogen by weight and has an C (Palmer & Cullis, H empirical composition of eight parts carbon to one part hydrogen, 8  10  1965). Theoretically, the soot is formed because of the growth of molecular precursors which consists of free radicals and ions through the process involving the conversion of  primary molecular species into relatively larger particles containing tens of thousands of atoms and far greater C/H ratio (Donnet, Bansal, & Wang, 1993). In its simplest form, this process is assumed to involve simultaneous polymerization and hydrogenation mechanism (Deviney & O’Grady 1976).  Figure 2.1 illustrates the typical steps involved in the soot formation mechanism. In general, the formation process of soot involves three stages: •  Inception of soot precursors and nucleation which involve the conversion of molecular system to a particulate system  •  Particle growth which involves the collisions and adhesions between large number of soot nuclei to form larger primary particles with typical dimension in the order of 10-30 urn  •  Agglomeration of these primary particles into chain-like clusters  The details on the soot particle inception and nucleation, particle growth and agglomeration processes will be further discussed below.  11  Soot Precursors  H 2 C  Primary Particle  Nuclei  Nucleation  ‘,  • vr• .i  •‘  I.  Coalescence  .  Surface Growth  Soot Aggregates  Aggregation  • .4.  Surface Growth  •• • $44 4  PAH  Figure 2.1 Schematic of the Soot Formation Process (Adapted from Svensson, 2005)  2.1.1 Soot Particle Inception and Nucleation The inception of the soot precursors is the least well understood stage in soot formation mechanism. However, the nucleation process is the most important stages of the whole soot formation process. Generally, the soot is formed in the fuel-rich region where the lack of oxygen prevents complete combustion of fuel. It was believed that the soot formation process usually begins with gas-phase precursors (Naydenova, 1.1., 2007; Oktem, B et al., 2005; Keller, Kovacs, & Homann, 2000; Vioila & Venkatnathan, 2006; Zhang, 2005). Several different theories have been proposed for the soot precursor inception which identify several possible soot precursors including C 2 (Gaydon & Fairbain. 1954), acetylene (Porter, 1955), polyacetylene (Homann & Wagner, 1967), polyynes (Frenklach et al., 1985), and ions (Calcote, 1981).  Among these theories, the most widely accepted is the polyaromatisation theory that assumes PAHs as the dominant molecular species for soot formation (Seinfeld & Pandis, 1998). Several PAH models have been proposed to explain the formation of the first  12  aromatic ring in the combustion process. One such model, given by Frenklach and Wang (1998), suggests that acetylene (C ) reacts with propargyl 3 H 2 F1 to form (C ) cyclopentadienyl 5 F1 radical. This C (C ) H radical will further react with methyl (CH 5 ) 3 radical which is usually produced from the fragmentation of larger hydrocarbon radicals to form benzene (C ) which is the base ring structure of PAH (Frenidach et al., 1985; H 6 Moskleva, Mebel, & Lin, 1996); it should be noted that other possible routes involving chemical reactions of radicals of smaller PAH with other PAll (Schuetz & Frenklach, 2002) or PAll radicals are also proposed  At some point after the precursor species is formed, it will evolve and grow to form larger aromatic ring. This growth is mostly due to the addition of carbon atoms following the hydrogen abstraction C H addition (HACA) mechanism (Frenklach et al., 1985; 2 Blanquart & Pitsch, 2007).  The individual precursor will keep growing to a certain size when the nucleation starts to occur. Frenklach et al. (1986) suggest that the soot nucleation process is basically a continuous PAH growth and coagulation process that produce nuclei in the form of large condensed PAll particles. Another suggested mechanism for soot nucleation includes the proposal that nucleation occurs due to the condensation of gaseous PAH species (Heywood, 1988) and the suggestion that the soot nucleation occurs due to a fast and irreversible polymerization of polyynes (Krestinin, 2000). In most cases, these tiny particles formed during the nucleation do not contribute significantly to the total soot mass, but do have a significant influence on the mass added later by providing sites for  13  surface growth (Svensson, 2005).  2.1.2 Particle Growth and Agglomeration Although the nucleation of the soot particle is the most important stage in the soot formation, it is important to recognize that only small amount of soot mass is produced by the nucleation process. Almost 90% of the total soot mass is produced as the result of particle growth (Tanner, Goodings, & Bohme, 1981). The particle growth usually involves two major processes: the coalescence of soot nuclei into larger particles and the surface growth by HACA. Both coalescence of soot nuclei and the surface growth will produce a particles with diameter of about 10 urn (Heywood, 1988). In addition, the deposition of organic materials on the surface of primary particles due to absorption and condensation of the organic fractions can also be observed during this growth process. The surface growth process and the deposition of carbonaceous materials on the particle surface are usually responsible for filling the void volume in the particles that are formed by the collisions and adhesions between the soot nuclei, and maintaining their spherical shape. At the end of the growth process, primary particles  —  often termed as spherules  with typical diameter of 15 urn to 30 urn are produced (Heywood, 1988).  As the particles keep growing, collision between larger particles starts to occur. For the collision between the particles with diameter larger than 10 urn, the effect of surface growth becomes relatively negligible since it is no longer fast enough to completely fill the larger void volume in the particles (Heywood, 1988). As the result, the product of collisions between large particles often loses its spherical shape (Mitchell & Frenklach,  14  1998). Moreover, this also explains why the individual primary particles in the exhaust still maintain the overall spherical geometry while the soot aggregate usually exhibit chain-like structure.  As the primary particles grow to a certain size, the agglomeration process occurs. During this process, the primary particles agglomerate into clusters that have chain-like structure. The typical soot aggregates from the exhaust of diesel engine are shown in Figure 2.2. During the agglomeration stage, the surface growth process still continues to merge the primary particles on the aggregate which causes the obliteration process to occur (Park & Rogak, 2003). The obliteration process is responsible for increasing the diameter of the primary particles, reducing the number of primary particles within the aggregate, as well as partly restoring the spherical geometry of the primary particles. The effect of the obliteration process will typically compete with the agglomeration process to decide the final number of the primary particles in the soot aggregate.  Figure 2.2 Soot Aggregates Produced by Diesel Engine  15  2.1.3 Soot Oxidation During and immediately after its formation and growth, oxidation of soot particles may occur which often results in a decrease of the soot concentration. Soot oxidation is classified as a heterogeneous (gas-solid) reaction, and generally happens when the soot particles diffuse to lean parts of the flame where the temperature and oxygen concentration are high. The soot particles are usually oxidized when they are burned in the presence of oxidant species such as molecular and atomic oxygen or the hydroxyl radical (Donnet, Bansal, & Wang, 1993). During the oxidation process, the soot particles will react with the oxidants to produce gaseous products such as carbon monoxide (CO) ). The rate of which the soot is oxidized depends on the diffusion rate 2 and dioxide (C0 and the kinetics rate of the reaction (Heywood, 1988). Up to 90% of the soot formed will be completely burnt during the oxidation process, and only the remaining portion of soot particles that survive the oxidation at the end of the combustion will be emitted into the exhaust (Heywood, 1988; Prakash, 1998).  2.1.4 Soot Restructuring After its growth and agglomeration, a soot aggregate often undergoes microstructural rearrangement, referred as soot restructuring. The soot restructuring process is usually characterized by the transformation of an initially elongated chain-like structure into more compact spherical globules with smaller size (di Staslo, 2001; Colbeck et al., 1990; Weingartner, Burthscher, & Baltensperger, 1997; Mikhailov et al., 1998). Evidences show that the soot restructuring occurs not only in flame (di Stasio, 2001), but also for  16  soot released in atmosphere (Colbeck et al., 1990; Mikhailov et al., 1998).  The aggregate restructuring occurring inside the flame is likely caused by local thermal enhancement (di Stasio, 2001) and Coulomb interactions (Onischuk et aL, 2003). On the other hand, the structural rearrangement of airborne soot particles in the atmosphere is likely induced by the capillary forces of condensing water vapor (Colbeck et al., 1990; Weingartner, Burtbscher, & Baltensperger, 1997; Mikhailov et al., 1998). As the water condensing on the surface of the soot aggregates, the deformation of the fractal structure occurs due to the surface tension (Mikhailov et al., 1998). This deformation process is usually accompanied by the breaking of interparticle bonds, which lead to a collapse of the aggregate into smaller fragments with a more compact structure (Mikhailov et al., 1998).  2.2 Quantifying Soot Aggregate Morphology and Microstructure Past research has shown that the scattering properties of soot are strongly affected by their morphology and microstructure. These effects have motivated extensive interest in the structure of soot. In this section, the studies on the size and shape of soot aggregate, as well as their microstructure will be reviewed.  2.2.1 Aggregate Size Past studies have revealed that primary particles of soot produced by combustion do not have separate existence, but rather the carbonaceous primary particles were agglomerated together to form clusters. (Lee et al., 2001; Clague et al., 1999 ; Mathis, Mohr, & Kaegi,  17  2005; Ishiguro, Takatori, & Akihama, 1997; Vander Wal, & Tomasek, 2004). The aggregate clusters are the most frequently encountered form of soot since they represent the most stable form of soot particles (Clague et al., 1999).  As the carbonaceous particulates are typically composed of agglomerated primary particles, the aggregates can be considered as the true representative of soot (Forrest & Witten, 1979). As a result, the quantification of the soot particles is commonly done based on the dimensions of the aggregate alone.  Several methods for quantifying aggregate size have been appeared in the literature. Most of the proposed methods typically involve the use of either the projected area or the perimeter of the aggregate. The most commonly used aggregate size classifier is an aggregate diameter, DA, which can be derived directly from the projected area alone. The aggregate diameter is determined from the 2-dimensional aggregate projection as a circle (Donnet, Bansal, & Wang, 1993), and can be calculated as follow: DA  2 (4A”  =—)  (2.1)  where A is the projected area of the soot aggregate.  For the characterization of the soot aggregates, the area equivalent diameter calculated from equation 2.1 is useful since it is closely related to the aggregate mobility diameter. The electrical mobility diameter is the diameter of a sphere with the same migration velocity in a constant electric field as the aggregate of interest (Flagan 2001). Past study done by Rogak, Flagan and Nguyen (1993) has shown that there is linear correlation  18  between the area equivalent diameter and the mobility equivalent size. Past study has shown that the mobility diameter of agglomerate (dm) should be approximately equal to its projected aggregate diameter DA for single charged particles (Rogak, Flagan, & Nguyen, 1993).  To date, a variety of methods are available for characterizing the soot aggregate size. In general, these methods can be classified as either real-time measurements or postsampling analysis.  Real-Time Sampling Techniques The real-time measurement techniques generally require diverting some flow of interest to a measuring instrument. The real-time field measurements are often used to determine distributions of the soot aggregate sizes using commercially available instruments, these include: Scanning Mobility Particle Sampler (SMPS), Passive Cavity Aerosol Spectrometer Probe (PCASP), and a PMS Forward Scattering Spectrometer Probe (FS SP). The measured size distribution is usually represented as a function of the mobility diameter. The detailed descriptions of each instrument are summarized in Table 2.1.  19  Table 2.1 Real-Time Sampling Techniques Instrument  Description  Size Range  SMPS  The particles are charged and sized based on their  2  (TSI, 2003)  mobility diameter  PCASP  An intrusive optical particle counter (OPC). The  (Strapp et al., 1992)  particles are sized based on the amount of light  —  320 nm  >150 nm  they scatter. FSSP  A non-intrusive OPC, the size of particles are  (Marshall et al., 2005)  determined based on the forward-scattered light  >  200 nm  from an incident laser  Among the instruments listed in Table 3.1, the SMPS is the most often used for research purposes due to its ability to measure nano particles. Most SMPS systems use electrical mobility detection technique by first charging the particle to a known charge distribution and then classifying them according to their mobility within tan electrical field (TSI, 2003). This detection technique allows SMPS to measure particles in smaller size range.  Recently, Laser Induced Incandescence (LII) and LS/E (light scattering/extinction) has been extensively studied for the purpose of measuring aggregate size (Kamimoto, Shimoto, & Kase, 2007; Schraml et al, 2000; Smaliwood et al., 2001). Although the LII method has shown a promise, it still suffer considerable uncertainties in the model and parameters (Krisbnan, Lin, & Faeth, 2001; Smaliwood et al., 2001). Furthermore, there is a probability that high-power laser emission used to the particles in the measurement volume can alter their morphology and internal structure. In addition, the light scattering method works based on an assumption of mono-disperses spherical particles that just touch one another. Therefore, the light scattering method can not be used to measure the 20  aggregates with compact structures since that assumption is compromised.  •  Post-Sampling Analysis Techniques  Generally, the post-sampling analysis works by collecting soot particles from the flow of interest on grid and subsequently analyzing them under electron microscope so that their morphology can be studied. There are two different types of electron microscopy that are most frequently used for soot characterization: transmission electron microscopy (TEM) and scanning electron microscopy (SEM). Both microcopy techniques can be used to image the aggregates. However, the TEM usually has advantage over the SEM due to the fact that it usually produces image with higher resolution (Gwaze et a!., 2005) which make the task of defining primary particle boundaries easier.  The transmission electron microscopy has a few limitations. The first drawback of the transmission microscope is that its ability to examine the soot aggregate is limited to a small area of the sample. Thus, lots of efforts and time will be needed for examining one single sample. Other disadvantage of the TEM is that the image analysis program needed for characterizing the soot aggregates from the TEM images often requires complex algorithm. Hence, developing the image processing program is often a difficult task.  Despite its drawbacks, TEM is still considered to be the most reliable method for determining the soot morphology. The reasons are: 1. TEM analysis can be used to directly quantify the aggregate morphology Aside from the projected area of the aggregate, the TEM analysis can provide other  21  important structural properties such as the projected aggregate perimeter, the size of the primary particles and the number of primary particles in the aggregate. 2. TEM allows qualitative observation of the aggregate The particular advantage of TEM is that the structures of the aggregates are readily visible. Hence, it will allow researchers to observe features that are not captured by quantitative measurements. 3. Contamination particles can be easily identified and excluded from TEM analysis Unlike real-time measurement techniques which are susceptible to interference caused by dust, water droplets and other small particles in the flow of interest, TEM analysis allows the researchers to exclude the contaminants from the measurement.  22.2 Aggregate Shape There is a considerable variation in the shape of the soot aggregates produced by combustion process, ranging from near perfect spheres to chain-like structures. Herd, McDonald, and Hess (1991) categorized the soot aggregate based on the aggregate/length width ratio. In general, the soot aggregate shapes were classified into four different types (Herd, McDonald, & Hess, 1991): 1. Spheroidal Spheroidal includes soot aggregates that has L/W ratio less than 1.5 2. Ellipsoidal The ellipsoidal has L/W ratio of 2.0-3.5 3. Linear Linear categories includes aggregates that have L/W ratio of>3.5. They usually have  22  elongated fiber-like main chains with limited number of branches and hence, show low irregularity 4. Branched Branched categories is basically the soot aggregates that have L/W ratio of>3.5 and are highly irregular.  Figure 2.3 shows the shape classifications for soot aggregates. It should be noted that the four shape categories mentioned above were tested with Hotellings T test (Kshirsager, 1972) and found that they are statistically different.  Spheroidal  Ellipsoidal  ..  . .•  S  Linear  Branched  ,(  1’’  ¼  Figure 2.3 Classification of Aggregate Shape (Adapted from Herd, McDonald, & Hess, 1991)  23  Soot aggregates emitted by engine rarely exhibit spheroidal shape and majority of them posses complex shape. The highly irregular nature of the soot aggregates makes the characterization of carbonaceous aerosols more complicated. Fortunately, past study has shown that the fractal dimension can accurately describe the aggregate structure (Meakin, 1983; Forrest & Witten, 1979; Meakin, 1998; Samson, Mulholland, & Gentry, 1987; Rogak & Flagan, 1992).  Several different types of fractal models have been used to characterize aggregate particles. The first method of fractal analysis on carbon black aggregates was based on the perimeter-area relationship (Kaye; 1984) to detennine a perimeter fractal as follows: (2.2) where P is the projected aggregate perimeter, A is the projected area and D is the perimeter fractal. The D is derived as the slope on al east square linear fit of the log P versus log A. However, the extent of the application of perimeter fractal is limited due to the fact that it could not fully describe the agglomerates that have highly acicular geometry (Donnet, Bansal, & Wang, 1993).  A more useful fractal model is the mass fractal which provides information on how an aggregate’s mass scales with its size (Forrest & Witten, 1979; Donnet, Bansal, & Wang, 1993). The general relation of mass fractal is as follows: (2.3) where M is the mass of the aggregate, R is the size scale and Df is the fractal dimension.  24  The mass fractal plays important role in the soot characterization since it has direct implication on the scattering properties of soot aggregates. Based on the Rayleigh-Deybe Gans polydisperse fractal aggregate (RDG/PFA) theory the light scattering of soot ,  aggregate can be accurately approximated as long as the fractal dimension, Df, used in the mass fractal analysis found from the relationship between the primary particle diameter (dr), the number of primary particles in the aggregate (N) and the radius of gyration (Rg);  Y  (R N=kfJ  (2.4)  is less than 2 (Sorensen, 2001, Shaddix et al., 2005). The kf in equation 2.4 represents the fractal prefactor. For soot aggregates produced by combustion, the stipulation can usually be satisfied since their Df value is usually about 1.8 (Megaridis & Dobbins, 1990; Farias, KOylti, & Carvalho,1996; Farias et al., 1995). Other literatures have shown that the Rayleigh-Deybe-Gans (RDG) scattering approximation can successfully estimate the scattering properties of fractal like soot aggregate within the experimental uncertainties (Muiholland, Bohren, & Fuller, 1994; Koylu, & Faeth, 1994).  2.2.3 Microstructure Soot microstructure, defined as the local spatial arrangements and orientation of the graphene layer segments, has significant influence on many important properties of soot, including the optical properties (Bond, & Bergstrom, 2006; Choi et al., 2001), electrical properties (Chen & Zhao, 2000), oxidation reactivity (Song et al., 2006), mechanical strength and modulus (Emmerich, 1995) and thermal expansion coefficient (Sato et al., 1988). These effects led to considerable efforts being spent on characterizing the soot  25  microstructure over the past decades.  Numerous attempts to characterize the microstructure of carbon black can be traced back as far as the 1920’s. However, early attempts to study the soot microstructure were largely limited to the X-ray diffraction analysis (Donnet & Voet, 1976) and hence, the detailed information on the internal organization of the soot particles was not fully obtained until high resolution transmission electron microscope is available.  The early study conducted using an X-ray diffraction shows that a soot aggregate is composed of bundles of parallel orientated graphitic planes and has intermediate structural form between graphite and nontridimensional structure (Donnet & Voet, 1976). This type of structure was defined as turbostratic (Biscoe & Warren, 1942) where the stacking of the graphene layers are still parallel, but the layers are rotated around by a series of arbitrary rotation in their planes (i.e., disoriented around the c-axis). The microstructural model based on the early X-ray diffraction studies shows random orientation of crystallites in the bulk of particle which are connected to one another either through single plane or by poorly organized carbon as illustrated in Figure 2.4.  IJJ’I  Figure 2.4 Early Model of Soot Microstructure (Adapted from Donnet, Bansal, & Wang, 1993)  26  The micro structural model of random crystallite orientation in soot particles derived from X-ray diffraction was proven to be inaccurate by the analysis done using high resolution transmission electron microscopy (HRTEM). Instead of randomly orientated, the HRTEM analysis reveals that the layer planes are orientated parallel to the surface (Donnet & Voet, 1976). Further study by Hailing and Heckman (1969) postulates that the building block of soot particles is the individual graphitic layers not crystallites and -  -  usually referred as paracrystalline layer structure (Donnet, Bansal, & Wang, 1993). The schematic of paracrystalline structure of soot is depicted in Figure 2.5. At present, this paracrystalline concept is the most widely accepted microstructural model for soot.  Figure 2.5 Paracrystalline Model of Soot (Adapted from Donnet, Bansal, & Wang, 1993)  In general, the form of soot microstructure can be accurately described as a concentric, onion-like structure with an outside armor of perturbed graphitic layers and an inner core of amorphous material. The graphitic-like platelets can be considered as highly disordered graphite since they are typically composed of 3-4 turbostratically stacked graphene layers (Homann, 1988; Lahaye & Prado, 1978). The amorphous materials usually consist of polycyclic aromatic compound, organic and inorganic materials in  iregu1ar arrangements (Sadezky et al., 2005). 27  On mesoscale level, a distinction between graphitic and amorphous carbon is made based on their carbon bonding. Generally, there are two types of bonding found in carbonaceous material: sp 3 (tetrahedral) and sp 2 (trigonal) bonds. The extensive arrays of 3 bonds are usually found in diamond; while sp sp 2 hybridized carbon is used to characterize pure graphite.  In diamond, the four valence electrons are assigned into a symmetric set of four tetrahedrally strong, isotopic a bonds as illustrated in Figure 2.6 (Shaddix et al., 2005; Bond & Bergstrom, 2006). The four bonds are usually of the same length and strength. Due to their orbital arrangements, material that contain sp 3 bonding are usually quite transparent to visible and infrared radiation (Shaddix et al., 2005). In graphite, three of the valence electrons are directed in a plane with an angle of 1200 between them to form a bonds in the plane as shown in Figure 2.7. The fourth valence electron forms a weak it bond perpendicular to the plane (Bond & Bergstrom, 2006; Lower, 2007). As the result, graphite is highly anisotropic material that has optical transparency in its axial direction and strong optical absorptivity perpendicular to its axis (Shaddix et al., 2005).  28  109°  Four sp 3 hybrid orbitals  (a)  120°  7/ 120° /  120°  Three sp 2 hybrid orbitals  (b) Figure 2.6 Tetrahedral (a) and Trigonal (b) Hybridizations (Adapted from Lower, 2007)  Combustion-generated soot usually contains a mixture of sp 3 and sp 2 bonds. On mesoscale level, the carbonaceous particles are composed of sp -bonded clusters that are 2 -sp bonded carbon. The sp surrounded by 3 2 hybridized carbon acts as a defect in the sp 3 matrix and tend to form planar graphitic microcrystallites diffused within it (Chhowalla et 29  al., 2000). The condensation of sp 3 bonds into sp -bonded layers are responsible for the 2 bent, paracrystalline graphene layers found in soot (Shaddix et a!., 2005).  Past researches fmd that the optical properties of soot are primarily governed by the extent of the sp 2 site (Robertson 1992; Chhowalla et al., 2000; Chena & Zhao, 2000) Increasing the extent of the sp 2 site causes a decrease in optical gap and increase in the absorption capability in soot (Bond & Bergstrom, 2006). Furthermore, the change in the  2 site will also affect the refractive index of the soot since the real part of the refractive sp index is dependent on the optical gap (Bouzerar et al., 2001).  In addition, part study also show that the sp 3 bonds contained in the carbon can transform into sp 2 bonds. This transformation process is usually called graphitization (Oberlin, 1984), and plays important role in determination of the optical properties of carbonaceous particles. The investigation done by Chhowalla et al. (2000) shows that the gradual ordering of sp 2 phase into graphitic clusters does indeed influence the optical gap and resistivity of the carbon. Consequently, it can be concluded that the quality and size of 2 phase present in carbon— often referred as the degree of graphitization sp  —  strongly  influence the optical properties of carbonaceous particles.  The soot particle microstructure amorphous carbon  —  —  specifically the ratio of crystalline graphite-like to  is highly dependent on the starting material and the synthesis  conditions such as temperature, time, fuel properties and fuel/oxygen ratio (Mathis, Moore, & Kaegi, 2005; Sadezky et al., 2005). Different carbonaceous particles may have  30  different mierostructure. Hence, it is important to be able to discriminate different type of soot. Several techniques that are routinely used to characterize the microstructure of carbon such as X-Ray Diffraction, high resolution transmission electron microscopy (HRTEM), scanning probe microscopy and Raman spectroscopy.  X-ray Diffraction X-ray diffraction is the most commonly used technique for characterizing carbon microstructure (Fujimoto, 2003; Sadezky et al., 2005; Dana et al., 2006; Sacanlon & Ebert, 1993). The X-ray diffraction can provide information about the crystallographic structure, chemical composition, and physical properties of materials by analyzing the scattered intensity of an x-ray beam passing through the srystal as a function of incident and scattered angle, polarization, and wavelength or energy (Cullity & Stock, 2001). The X-ray diffraction is really sensitive to the crystalline (long range order) structure of the carbon and hence, can directly characterize the average stacking height of carbon lamella and the average size (diameter) of the layer planes (Donnet, Bansal, & Wang, 1993).  However, X-ray diffraction has two major disadvantages. First of all, X-ray diffraction  can not measure some aspects of carbon microstructure such as the curvature of the carbon layers and the linearity of the layer planes. In addition, other aspects of carbon microstructure such as a mean lattice plane size can not be characterized accurately due to low resolution accuracy of the diffraction method (Vander Wal et a!., 2004). Secondly, the X-ray diffraction measurement does not compensate for layer bending and thus, it often underestimates the value of the average diameter of the layer planes.  31  High resolution transmission electron microscopy (HRTEM) HRTEM technique is often used to characterize the carbonaceous particles due to its capability of providing structural information at atomic level (Chen et al., 2005). In contrast to the diffraction method, HRTEM can quantify almost all aspects of carbon microstructure including the curvature (tortuosity) of the lamella, fringe length and separation. Moreover, the HRTEM analysis provides direct visual observation and hence, the subtle differences in the microstructure can be observed right away (Vander Wal et al., 2004).  HRTEM also suffers some drawbacks such as HRTEM imaging can only examine localized, micro-quantities of sample, Hence, it lacks the ability for quick discrimination of different type of soot. Another disadvantage of the HRTEM method is that it is really difficult to acquire high-resolution images of soot and consequently, the HRTEM analysis can be time consuming. In addition, to measure some the properties of the carbon microstructure such as fringe length and tortuosity, image processing programs is needed for complete analysis. Developing this image processing algorithm can be a challenging task since it is really difficult to extract a set of identifiable fringes from HRETM images (Shim, Hurt, & Yang, 2000).  Scanning probe microscopy The scanning probe microscopy (SPM) has been employed to analyze the surface of carbonaceous particles with the resolution of single atoms (Chen et al., 2005, Barone,  32  D’Alessio, & D’Anna, 2003; Lehmpuhl et al., 1999; Ravier et al., 2001). The most commonly used scanning probe microscopy techniques for soot characterization are the scanning tunneling microscopy (STM) and the more advanced atomic force microscopy (AFM). Both STM and AFM were often used to determine the surface topography of carbon black since they have unparallel ability to measure small local differences on the carbon surface. This ability was caused by the fact that the tunneling current of the STM is exponentially proportional to the gap spacing og the carbon and hence, is really sensitive to even the smallest changes in gap spacing (Donnet, Bansal, & Wang, 1993).  Despite its accuracy in analyzing the surface of soot, the SPM method also two major disadvantages. The first disadvantage of SPM is that it can only produce small images which make the task of analyzing the image more difficult, The second main drawback is that the SPM requires the surface of the sample specimen to be flat. A variation of the height of the sample surface larger than 10 tm can possibly affect the measurement.  Raman spectroscopy In recent years, Raman spectroscopy has become the preferred method for characterizing the microstructure of carbonaceous particles. Rosen and Novakov (1978) are the first to use Raman spectroseopy to analyze the microstructtire of diesel soot. The results of their investigation prove the presence of graphitic structure in the soot produced by a diesel engine. Rosen and Novakov’s investigation has been followed up by others (Dippel & Heintzenberg, 1999; Mernag, Cooney, & Johnson, 1984, Cuesta et aL, 1994; Jahwari, Roid, & Casado, 1995; Ivleva et aL, 2007). Most of these studies used Raman  33  spectroscopy to distinguish diesel engine soot and other industrial soots based on their degree of graphitization.  Unlike X-Ray diffraction, Raman spectroscopy is sensitive not only to crystal structure (long-range order), but also to molecular structure (short-range order). Hence, Raman spectroscopy is definitely a better choice for characterizing highly disordered material such as soot since it is sensitive to the degree of the structural disordered. Furthermore, compared to HRTEM analysis, the Raman spectroscopy offers a capability to examine larger area of the sample and therefore, can be easily used to discriminate different types  of soot.  34  3O  3.1  DIESEL SOOT MORPHOLOGY  Introduction  The first objective of this study is to characterize the engine-emitted soot aggregate morphology, emphasizing the fractal properties. To do this, the transmission electron microscopy (TEM) was employed to image the soot aggregates and derive the fractal properties of the aggregates. The TEM was the method of choice since it remains the most accurate means to measure the morphological properties of fractal-like aggregates. A Matlab-based image processor was developed to extract the geometrical properties from the TEM images including the projected soot aggregate length, width, area and perimeter, as well as the size distribution of the primary particles.  In this study, the fractal dimensions (Df) were computed using two different approaches. The first approach utilized a method based on the projected aggregate dimensions, which involved the use of the projected length (L) and width (W) of the aggregate. The second approach employed the pair correlation function for estimating the fractal dimension of the aggregate. The fractal analysis results from both methods will be compared and discussed in this chapter.  35  3.2  Experimental Setup  3.2.1 Soot Samples Samples from Light-Duty Diesel Engine A 2001 four-cylinder Volkswagen 1 .9L turbocharged direct injection (TDI) engine was used to sample different particulates from the exhaust manifold immediately downstream of the exhaust valves. The specifications of the VW TDI engine used in this study are summarized in Table 3.1.  Table 3.1 Volkswagen TDI Engine Specifications Parameters  Values  Compression Ratio  19.5:1  Displacement  1.896 L  Cylinder Bore  79.5 mm  Piston Stroke  95.5 mm  Max Power  60kW  Max Torque  @ 3300 rpm 205 N.m @ 1800 rpm  Charge Pressure  0.92 bar  The exhaust stream was sampled downstream of the turbocharger outlet via two-stage dilution system. In this dilution system, the engine exhaust was mixed with a clean diluent (nitrogen in this case) at the first stage of dilution (i.e., the primary dilution stage). Control of the primary dilution stage was achieved by adjusting the nitrogen mass flow rate. The primary dilution ratio was determined by comparing the CO 2 concentration in the exhaust gas and diluted flow; and a correction was made for the background CO 2 concentration from the bottled N . During the tests, the primary dilution system was 2 operated at a dilution ratio of 27-28:1; while the temperature was maintained at about 52°  36  C by preheating the line at the primary dilution stage. A portion of the diluted exhaust was passed through the secondary dilution stage, where it was further diluted by air via a mini-dilution timnel. The secondary dilution ratio was set at approximately 33:1 for all tests.  For the transmission electron microscopy (TEM) study, the soot samples were captured after the primary stage of dilution using thermophoretic sampling techniques from the raw exhaust and deposited onto holey carbon grids. The design of the thermophoretic sampler used in this study can be found in Appendix A. The sampling objective was to collect a single layer of particles so that the structure could easily be examined with the microscope. In order to achieve this objective, the sample time was varied according to the particulate emission rates. Samples times ranged from 30 seconds to 4 minutes.  The size distribution measurements were made after the secondary dilution stage to ensure that the particle concentration of the sample flow did not surpass the maximum limit for the scanning mobility particle sizer (model 3936, TSI Inc.), and to prevent saturation in the condensation particle counter (CPC) (model 3022, TSI Inc.). The particle size distributions in the exhaust samples were measured simultaneously using SMPS with two different differential mobility analyzers (DMA): a nano-DMA (model  3025, TSI Inc.) and a long-DMA (model 3081, TSI Inc.). Data was exported using Version 8.0.0 of the TSI Aerosol Instrument Manager software for both instruments using diffusion and multiple charge correction.  37  To select the test engine operation conditions, preliminary testing was conducted. A range of engine operating conditions were evaluated to select 6 modes which would give a range of carbonaceous fractions of the particulate matters. The soot samples were collected with the 37-mm diameter Quartz fiber filters (pre-cleanecl) for PM 5 mass and . 2 5 organic and elemental carbon content measurement. . 2 PM  Organic carbon (OC) and elemental carbon (BC) were resolved using the NIOSH 5040 method, which employs evolved gas analyzer (EGA) technique by thermal optical analyzer. The schematic of the thermal optical analyzer typically used in the NIOSH 5040 method is shown in Figure 3.1 and its principles of operation are discussed elsewhere (Johnson et al., 1981; NIOSH, 1999).  Fm He  methanator 4 He/CH  photodiode  thermocouple aco oxidizer  lie 2 He/0 Figure 3.1 Thermal Optical Analyzer Design (Adapted from NIOSH, 1999)  38  Each engine mode was tested at least twice and the results shown here are averages of repeat measurements. Figure 3.2 summarize the measurement results. The fraction of 25 mass contributed by carbon is shown in Figure 3.3. In general, the carbonaceous PM content of the PM ranges from 58% to 86% with an average of around 69%. It is unclear why the carbonaceous material does not account for a larger fraction of the mass especially since it is expected that the carbonaceous fraction usually accounts for 95% of the mass. It is unlikely that this fraction is accounted for by sulfates because the fuel sulfur content was 15 ppm. Sampling errors such as leaks in the filter pack may be the cause. However, this should not affect the relative distribution between organic and elemental carbon, so the results can be used for the purpose of selecting modes that produce a range of PM compositions.  Variation of PM Composition by Engine Operating Condition 1oooo  1  1 903  j  59 ;119 38  131159  119’ 96  163’ 19  38 53  76 1142  1603 1603 1896 1910 2399 24 2603 2601 3743 3746 3742 3745 3744  TorquelSpeed  Figure 3.2 Variations of PM Composition by Engine Operating Condition  39  Carbonaceous Fraction of PM 100% 80%  L  —______  i1[[[[I:[1 IEIt[ii 159i119  903  19  38131I59I119I96I163  38  5376  1421  1603 160311896 1910 2399 2404 2603 2601 3743 3746 3742 3745 3744 Torque!Speed  Figure 3.3 Carbonaceous Fractions of Particulate Matters  The final six modes selected for this study were listed in Table 3.2. These modes are  selected since they gave a suitable range of organic/elemental carbon compositions as shown in Figure 3.4. Table 3.2 Engine Test Modes 1  2  3  4  5  6  Speed (RPM)  3750  3750  3750  2600  1900  1900  Torque (Nm)  76  38  19  162  154  38  Power (kW)  30  15  8  44  31  7  Mode  40  OCITC Ratio (pink modes selected for final test sequence) 40%  36%  35%  32%  30%  25%  21%21%I  20%  1 4%15%  119 1603  59T163  119  2601  I; 2404  2399  16%  15%  96J38T59T131 2603  1896  1603  76  1910 3745  1  142T38  903  3744 3746  19 3742  Figure 3.4 OC/TC Ratios of Particulate Matters  For this study, the engine was tested with and without an AP Merit Catalytic Converter after-treatment system running on two different test fuels: ultra-low sulfur diesel (ULSD) with 15 ppm sulfur content and B20 blends (i.e., a blend of 20 wt% Canola methyl esters in ultra-low sulfur diesel). Table 3.3 lists the key properties of the test fuels.  Table 3.3 Properties of Test Fuels Parameters  ULSD  B20  Sulfur Content (wt ppm)  15  8  Oxygen Content (wt %)  0  2.4  (%)  25.5  18  Cetane Number  50  65  Density (glcm ) 3  0.837  0.876  Aromatic  41  3743  Soot Samples from Heavy-Duty Engine The second set of the soot samples was produced by a Cummins ISX series heavy-duty six-cylinder, four stroke, direct-injection engine This engine has been modified to operate on one cylinder as described by MeTaggart-Cowan et al. (2003). The specifications of the  Cummins ISX engine used in this study are summarized in Table 3.4.  Table 3.4 Cummins ISX Engine Specifications Values  Parameters Compression Ratio  16.7:1  Displacement  2.491 L  Cylinder Bore  137 mm  Piston Stroke  169 mm  Rated Speed  1800 rpm  The Westport Innovations Inc.’s direct injection pilot-ignited natural-gas fuelling system was used in the engine. In this system, almost all the energy is supplied from the combustion of the natural gas which is injected into the cylinder before the piston arrives at top dead center. A small amount of diesel pilot, injected shortly before the natural gas,  was used to initiate combustion. For this research, the engine was operated at three different modes as listed in Table 3.5.  Table 3.5 Test Modes for Cummins ISX Engine Mode  1  2  3  Speed (RPM)  1200  1200  1200  Torque (Nm)  247  207  168.7  GIMEP (Bar)  8.5  10.4  12.5  42  Figure 3.5 shows a schematic diagram of the dilution system used for PM sampling and measurement. A small portion of the raw exhaust from the engine was drawn into the dilution system by a pump and mass flow controller. This system consists of primary and secondary dilution stage. In the primary dilution stage, the raw exhaust was mixed with a diluent (nitrogen). The primary dilution stage was controlled by setting the nitrogen mass flow rate, The ratio of the primary dilution was determined by measuring the CO 2 concentration in the exhaust gas, the diluted flow and the bottled N 2 flow, For all tests, the primary dilution ratio was set at around 12:1,  To Eithoust Fan  To Ethaust Duct  Filter Holder  LEGEND Pump MEM Mass Flow Meter  [XI Valve Needle Valve Thermophoretic  Au  Sampler  Flow Sensor  Filter P  Pressure Sensor  T  Temperature Sensor  To Ethaust Fan NDMA  cPc To Ambient  Figure 3.5 Dilution System  43  The sample line has multiple ports where a portion of the diluted exhaust sample is drawn into measurement instruments. Similar to the setup of the dilution system in the VW engine, a small portion of diluted sample was passed through the secondary dilution stage, where it was further diluted by filtered air. The secondary dilution ratio was determined by the flow measurement. The secondary dilution ratio can be determined by knowing the volumetric flow rate of the dilution air and the flow rates drawn into the measurement instruments. The secondary dilution ratio was set at approximately 10:1 for all tests, resulting in a total dilution ratio of approximately 120:1 for the sample flow going to the differential mobility analyzers (DMA).  After the primary stage of dilution, the soot samples were collected on holey carbon grids using thermophoretic sampling techniques for transmission electron microscopy (TEM) analysis. The soot particles were also extracted using the thermophoretic sampler described in Appendix A.  The size distribution measurements were taken after the secondary dilution stage to ensure that the particle concentration in the sample flow was within the recommended range for the scanning mobility particle sizer (model 3936, TSI Inc.) The particle size distributions in the exhaust samples were determined using SMPS with a nano differential mobility analyzers (model 3025, TSI Inc.), combined with the condensation particle counter (CPC) (model 3025, TSI Inc.). Data was recorded with the TSI Aerosol Instrument Manager software (version 8.0) using multiple charge correction.  44  3.2.2 Transmission Electron Microscope The TEM analysis was performed on the soot samples using a Hitachi H7600 microscope operating at 120kV with a nominal resolution of 0.2 nm. The instrument was controlled with Quartz PCI software. TEM images of thirty-five randomly selected soot agglomerates were taken for each grid. To avoid biased samples, the following sampling method was used: 1. Mark nine particular spots on the sample grid -displayed as circular platform with a radius of 1000 by the TEM software interface- as illustrated in Figure 3.6. 1000 -  *  500  •  / —1000  —500  •  1CJ0  500  —500  /  • /  Figure 3.6 Software Interface of the TEM Sample Grid  2. Go to one of the marked point and find the nearest TEM grid-square. 3. Move to the upper left corner of the TEM grid square and increase magnification until the field of view is about 1 micron across. 4. Move diagonally across the square until first particle is found. Take an image at the appropriate magnification.  45  5. Decrease the magnification to 1 micron field again and continue moving to next particle, and take the photograph. 6. Repeat step 4 to 5 until four images are taken from that particular grid. 7. Go to another marked point and repeat step 3 to 6.  The images were collected under optimal focus condition at a typical magnification of 400K-600K. This range of magnification was chosen to make the task of defining the primary particle boundaries easier. Furthermore, only unsupported agglomerates hanging over the gaps between the holey carbon coatings were analyzed to avoid interference from the holey carbon film  3.3  MATLAB Image Processing  In order to conduct the morphological characterization of the soot aggregate imaged by the TEM, an image processor needed to be developed. Several commercial software packages are available for image processing such as MATLAB (The Mathworks), which is used in this study that is. MATLAB is a computer program that uses high-level language, and an interactive environment that allows researchers to perform computationally intensive tasks much faster than traditional programs, such as C, C++, and Fortran. MATLAB is ideal for image processing since it works using a matrix oriented language and hence, each TEM image can be represented by a matrix with each element corresponding to a pixel of that image. Moreover, MATLAB also offers specialized toolboxes such as the Image Processing Toolbox and Statistical Toolbox which can be useful for processing and analyzing TEM images.  46  In this study, a program called Semi-Automatic Image Processor (SAIP) was developed in MATLAB with the help of Dorian Tsai. The main purpose of the SAIP is to extract the geometrical properties of the soot aggregate which include the projected agglomerate Length (L) and width (W), the projected aggregate area and perimeter, the number of primary particles in the aggregate (N), the mean primary particles mean diameter (dr) and the length (1) and width (w) of each primary particle as illustrated in Figure 3.7. In addition, the SAIP also computes the density-density correlation and pair correlation of each aggregate.  Figure 3.7 Schematic of Geometrical Properties of Soot Aggregate  The information obtained using the SAIP is saved in a text file format to allow thtuie 47  access. Along with this text file, the on-line database also contains several images of the aggregate including the original TEM image, the binary image of the aggregate, the image of the aggregate edge, the image showing the length and width axis, and the plots of density-density and pair correlation functions. A sample of the files stored in the database can be found in appendix B.  A modular approach was employed for the programming task in MATLAB as shown in Figure 3.8. A module is a component of a larger system that operates within that system independently from the operation of the other components (Etter & Kuncicky, 1996). In  this case, the SAIP program has been divided into different functions. For example, the binary function was designed to process the images and obtain the binary images of the agglomerate; while the calculations function was designed only to extract data. The modular approach was beneficial, because the program was easier to understand, easier to debug, and more manageable. The features of the functions involved in the SAIP will be briefly discussed below. The detailed codes of the SAIP can be found in appendix C.  48  Main Program  1. Binary  2. CalculatIons  3. Diary  Scale  Primary Particles  Area  Perimeter  Length & Width  IR_Coefficient  Area_R_Whole  Density-Density Correlation  Pair Correlation Function  Figure 3.8 Modular Diagram of the SAIP Program 3.3.1  Main Program  In the SAIP, the main program serves as the governing function that accepts the user inputs and divides the program into three major tasks: image processing, extracting the necessary information from the processed TEM image, and saving all the data in a text file.  3.3.2  Function Binary  The function ‘binary’ is designed for image processing purposes, and is essentially used to convert the digital image from the TEM into binary image. The first step in the image  49  processing is cropping the TEM image to select the desired particle. The cropped image is then filtered to reduce the background noise and improve the contrast between the aggregate and the background. The original TEM image and cropped image are shown in Figure 3.9.  (a) Figure 3.9 The Original (a), and Cropped and Filtered (b) Image of the Aggregate  The filtered image is then transformed into binary image by setting a threshold brightness value. The image of the aggregate edge can be easily obtained from the binary image. Lastly, the image of the aggregate edge is super-imposed onto the cropped image to make sure that the produced binary image is the accurate representative of the aggregate. The binary image, the image of the aggregate edge and the super-imposed image are shown in Figure 3.10  50  Figure 3.10 The Binary Image (a), Edge Image (b), and Final Imposed Image (c)  3.3.3  Function Calculations  The function ‘calculations’ supervises several sub-functions that are used to determine the morphological properties of the soot aggregate. The function simply manages all the sub-functions by calling them in a specified sequence, and passing the correct variables to the functions.  The sub-function ‘scale’ determines the scale factor, which is the conversion factor between pixel and the real length, from the scale bar obtained from the original TEM image. The sub-function area and perimeter measure the projected area and perimeter of the aggregate from the binary image, respectively.  The sub-function ‘length-width’ automatically determines the length and width of the agglomerate. In addition, this function also generates the image of the length and width drawn over the aggregate edge as shown in Figure 3.11.  51  Figure 3.11 The Length (Blue) and Width (Red) of the Aggregate  The sub-function ‘primary particles’ measures the length and width of each primary particle, as well as the mean diameter of the primary particle. This function relies on the user for differentiating the primary particles in the aggregate. Figure 3.12 illustrates the image of the length and width distribution of the primary particles.  Figure 3.12 The Length (Blue) and Width (Red) of the Primary Particles  The sub-function ‘tr coefficient’ is used to obtain the threshold-radius coefficient, which is defined as the relationship between the threshold value and the radius of the circle. This threshold-radius coefficient allows one to control the radius of the circle by changing the threshold value. The sub-function ‘areaR_whole’ calculates the area of a black ring of an outer radius r, and a thickness of dr = 1, and returns it in matrix fcrm, 52  with an increasing r. Both the ‘tr_coefficient’ and ‘area_R_whole’ provide the necessary components for calculating the density-density and pair correlation functions.  The sub-function ‘density-density correlation’ determines the density-density correlation function of the agglomerate. The density-density correlation function, C(r), represents the correlation between the density p (F’) at a point in the aggregate and the neighboring densities p (F  +  F’). The density correlation is given by the equation: (3.1)  where N is the number of primary particles in the aggregate and <>Q(F) is the average over all the angles of r. In general, this function calculates the density correlation function by counting the number of pairs of particles located on a ring with an outer radius r and an inner radius (r-1) within the aggregate, and dividing this number by the value obtained using sub-function ‘area_R_whole’. The plot of the calculated density density correlation is also provided as shown in Figure 3.13.  53  Density-Density Correlation Function  0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0  0  10  20  3J  40  50  60  70  Figure 3.13 The Density-Density Correlation Function of the Aggregate  Finally, the sub-function ‘pair-correlation’ determines the pair correlation function of the aggregate. The pair correlation function, C(r), basically represents the probability of  finding two particles within a given radius r. In general, this function calculates the pair correlation function by determining the probability of finding another particle at a neighboring point from a reference particle within an infinitesimal region on the particle. The plot of the calculated density-density correlation is also provided as shown in Figure 3.14.  54  Pair Correlation Function  (-)  Figure 3.14 The Pair Correlation Function of the Aggregate  The pair correlation function can be used to verify the density-density correlation function since both functions are essentially similar. In the SAIP, the density-density correlation calculates the areas of many circles within the particle to obtain the density correlation plot, while the pair correlation calculates the distances between each pixel that are located on the particle with reference to every other pixel, and plot the occurrence frequency of every single distance to find the pair correlation function. When plotted on the same graph, the density-density correlation and the pair correlation functions generally have positive correlation as shown in Figure 3.15. The discrepancy between the density-density correlation and the pair correlation functions is caused by the fact that the distances computed in the function pair-correlation are often rounded to the nearest decimal point.  55  The Pair Correlation and Density-Density Correlation Functions  C-)  10  20  30  40  50  60  70  80  Figure 3.15 The Pair and Density-Density Correlation Functions of the Aggregate 3.3.4  Function Diary  The function ‘diary’ is used to save the information determined by the function ‘calculations’ into a text file. This text file also includes administrative information such as the engine mode, engine fuel type, engine test date, TEM session date, magnification, the file name of the original TEM image, and the number of aggregates in the original image. The text file format is chosen so that it can be read by MATLAB for future analysis. A sample of the database text file can be found in appendix B.  3.4  Fractal Analysis  In this study, the morphological information obtained from the MATLAB program was used to derive the fractal dimension of the soot aggregates. Two different methOds of  56  fractal analysis were employed: the projected aggregate dimensions method and pair correlation method.  3.4.1 Projected Aggregate Dimensions Method In general, the mass fractal dimension of soot aggregates (Df) follows the following relationship: DJ  (R N=kjJ  (3.2)  where N is the number of the primary particles in an aggregate, Rg is the radius gyration of the aggregate and d is the mean diameter of the primary particle and kf is a constant prefactor.  However, the information needed for direct evaluation of equation (3.2) is often not available. Hence, a simplified method based on the projected aggregate dimensions is often used instead. This simplified approach involves the relationship between the number of the primary particles in an aggregate and the projected aggregate length, L, and width, W, as follows (Tence, Chevalier, & Jullien, 1986 ; Jullien & Botet, 1987; Koylu & Faeth, 1992 ; KoylU et al., 1995): N=  (3.3)  where kw is the constant fractal prefactor. In the equation (3.3), the (LW)” 2 acts as surrogates for Rg. The fractal dimension is derived as the slope of the least square linear fit of the plot of log N versus log ((LW)” 2 / d) as illustrated in Figure 3.16.  57  102  101  10°  +  620, Mode 1 No Oxidation Catalyst N  =  1.295 ((L% /d) 1 .714  _I 10 10°  (LVV’/dp  Figure 3.16 Illustration of the Projected Aggregate Dimension Method  3.4.2 Pair Correlation Method The pair correlation function is the weighted average of the particle density at a distance r from a particle. For a finite number of primary particles N, the correlation function is given as follow (Cleary, Samson, & Gentry, 1990): C(r)= 2NP(nJ +r)p(r)  (3.4)  where r is the location of thejth primary particle. The pair correlation function for a fractal aggregate has the following dependence, C(r)r’  (3.5)  where DE is the Euclidean dimension. For projected aggregate, DE is equal to 2 in equation 3.5. C(r) obeys the power-law scaling for r between the upper and lower cutoff radius. For example, the lower cutoff radius would be somewhat larger than the average  58  radius of the primary particles, while the upper cutoff radius would be the radius of gyration (Rogak & Flagan, 1992). The fractal dimension is then defined as the slope of the least square linear fit of the plot of log C(r) versus log r as illustrated in Figure 3.17.  101  4  10_5 10•  I  10  10°  102  Figure 3.17 Illustration of the Pair Correlation Method  3.5  Results and Discussion  3.5.1 Particle Size Measurement Soot SamDles from VW TDI Engine The geometric mean mobility diameters of the soot produced by a VW TDI engine were measured using the SMPS with both nano-DMA and long-DMA. The nano-DMA measured the particle in the size range of 4.4  —  the particle distribution in the size range of 13  168 rim, while the long-DMA evaluated  —  833 rim. Table 3.6 summarizes the  geometric mean diameters as calculated by the TSI Aerosol Instrument Manager  59  software.  Table 3.6 Geometric Mean Diameter for Soot from VW Engine Mode I  Mode 2  Mode 3  Mode 4  Mode 5  Mode 6  ULSD, Without Oxidation Catalyst Long-DMA(nm)  61.4  57.6  55.8  66.9  69.3  56.3  Nano-DMA(nm)  39.1  40.2  39.5  39.1  39.4  35.9  Long-DMA (nm)  58.9  56.6  55.2  66.6  69.3  55.6  Nano-DMA (nm)  38.5  37.8  37.8  37.9  38.7  35  Long-DMA (nm)  60.9  55.7  55.2  66.7  69.2  56.2  Nano-DMA (nm)  38.1  38.7  38.5  37.2  38.1  35.7  Long-DMA (nm)  58.2  56.4  54.2  66.1  69.1  55.7  Nano-DMA(nm)  36.4  37.9  36.3  37.8  38.1  35.1  B20, Without Oxidation Catalyst  ULSD, With Oxidation Catalyst  B20, With Oxidation Catalyst  In general, the presence of an oxidation catalyst tends to decrease the geometric diameter. Similarly, biodiesel blend also tends to decrease the geometric mean diameter. This happened because biodiesel contains oxygen, and past studies have shown that the presence of oxygen reduces the soot formation (Sakurai et al., 2003; Jung, Kittelson, & Zachariah, 2006) via soot oxidation process. Furthermore, the geometric mean diameters from the nano-DMA size distribution are smaller than those obtained using the longDMA, because of the difference in the size-range. The long DMA lower size limit did not permit the full shape of the distribution to be obtained and thus, it overestimated the mean diameters. On the other hand, the nano-DMA had an upper size limit and consequently, underestimated the mean diameters.  60  The geometric mean mobility diameters of the soot aggregates measured by the SMPS were compared with the projected aggregate diameters (dA) obtained from the TEM analysis in Table 33. Each projected aggregate diameter from the TEM measurement represents the median diameter (the diameter for which exactly one-half of the aggregates  are smaller and one-half are larger) from thirty-five aggregates. The median diameter is used in the comparison due to the fact that it is closely related to the geometric mean diameter (Seinfeld, 1986). The upper and lower numbers in the bracket correspond to the th 60  th and 40 percentile, respectively, of the projected aggregate diameters from TEM  measurement.  61  Table 3.7 Mobility Diameters and Projected Aggregate Diameters for VW Soot Mode I  Mode 2  Mode 3  Mode 4  Mode 5  Mode 6  61.4  57.6  55.8  66.9  69.3 55.8 53.1 (43.6)  56.3  ULSD, No Oxidation Catalyst Long-DMA (rim)  50.1  DA from TEM Analysis (nm) Nano-DMA (rim)  46.7(39.6)  45.8  39.1  51.3  ()  53.2  46.2(42.5)  40.2  39.5  58.9 52.1 46.3 (39.2.  56.6 49.7 449 (38.2.’  46.1(395)  38.5  37.8  60.9  55.7  48.4  54.0 (41.3)  47.2  53.4  39.1.  39.4  36.5  66.6 477 (.s)  69.3 56.8 52.2  55.6 51.6 46.2 (38.7  37.8  37.9  38.7  36  55.2  67.7  69.2  56.2  820, No Oxidation Catalyst Long-DMA (nm) DA from TEM Analysis (nm) Nano-DMA (nm)  .  55.2 48.4  54.1  ULSD, Oxidation Catalyst Long-DMA (nm) DA from TEM Analysis (rim) Nano-DMA (rim)  50.3 (38.6)  50.1  48.7  52.9  55.6  44.2(40.2)  44.1(393)  48.6(42.7)  51.9(41.3)  37.1  37.7  36.5  37.2  38.1  58.2 51.6 46.2 (37.2)  56.4  54.2  69.1  44.8(38.9)  66.1 52.5 47.4 (40.8)  56.5 498(41.8)  36.4  37.5  37.8  38.1  46.3  52.2  35.7  B20, Oxidation Catalyst Long-DMA (nm) DA from TEM Analysis (rim) Nano-DMA (nrn)  48.5  44.3  52.1 (36.5)  36.3  55.7 49.7 46.1 (38.3 36.1  As expected, the mean values of the projected aggregate diameters from the TEM measurements are larger than the mean mobility diameters from the nano-DMA measurements, but lower than the mean mobility diameters from the long-DMA measurements. The projected aggregate diameters from the TEM measurements confirmed that the presences of oxidation catalyst and biodiesel blend indeed tend to decrease the diameters of the particles.  The average primary particle diameters (dr) in the soot aggregates were also obtained from the TEM analysis. Table 3.8 summarizes the mean values of the d, and their 95% confidence interval. The confidence intervals were calculated from the t distribution. The  62  primary particle diameters were measured in the range of 35.9  —  39.3 nm across all  engine conditions. It was observed that the presence of the biodiesel generally tends to decrease the mean value of the primary particle diameter.  Table 3.8 Mean Diameters of Primary Particles for VW Soot Mode  No After Treatment  Catalytic Converter  RPM  Torque  Fuel Type  (N.m)  ULSD  820  1  3750  76.4  36.8 ± 4.2  35.9 ± 4.7  2  3750  38.2  37.7 ± 3.8  37.1 ± 4.3  3  3750  19.1  38.8 ± 4.4  38.3 ± 5.5  4  2600  163.5  39.3 ± 5.3  38.9 ± 5.2  5  1900  155  37.6±6.1  36.9±4.5  6  1900  37.7  37.2 ± 4.9  37.0 ± 5.4  1  3750  76.4  36.4 * 3.7  35.7 ± 5.6  2  3750  38.2  37.5 ± 4.6  37.4 ± 4.8  3  3750  19.1  38.6 ± 2.8  37.9 ± 5.6  4  2600  163.5  39.1 ± 4.2  38.4 ± 4.5  5  1900  155  37.3 ± 3.8  37.1 ± 4.3  6  1900  37.7  37.1 ± 4.1  37.1 ± 4.7  Soot Samnles from Cummins ISX Engine The particle size distributions of the soot produced by the Cummins ISX engine were measured using the SMPS with nano-DMA. The particle size distributions are represented as the function of mobility diameter as shown in Figure 3.l. The particle number concentrations have been corrected by the total (primary and secondary) dilution.  63  Particle Number Concentration of Soot 8.OE+06 7.OE+06 6.OE+06  4’  5.OE+06  • Mode 1: 1200 rpm, 247 N.m  . Mode 2: 1200 rpm, 207 N.m Mode 3: 1200 rpm, 168.7 N.m  .j 4.OE+06 .2  I  ! 3.OE+06  z  — ‘I  2.OE+06 -  1.OE+06 0.OE+00 1  10  100  1000  dm (nm)  Figure 3.18 Particle Size Distributions for Different Engine Modes  There are two different modes in which the soot particles are formed: an accumulation mode and nuclei mode. The accumulation mode is typically comprised of large agglomerates which are the results of particle growth due to coagulation during combustion. The nuclei mode usually consists of particles with diameters less than 20 urn that are formed by nucleation. As observed from Figure 3.18, both the nuclei and accumulation mode particles were present.  From Figure 3.18, the particle number concentrations were affected by engine loads. The total number concentration of particles increased with engine load. The particle number concentration, however, does not vary greatly in the 4—25 urn diameter range. On the contrary, Figure 3.17 shows a noticeable increase of the particle number concentration in the accumulation mode as the engine load was increased. This trend can be attributed to  64  the increase of fuel consumption and combustion emissions with the increase in engine load.  The geometric mean diameters of the soot aggregates measured by the SMPS were compared with the projected aggregate diameters (dA) obtained from the TEM analysis in Table 3.9. Similar as before, the projected aggregate diameters from the TEM measurement represent the median diameters, and he upper and lower numbers in the bracket correspond to the  th 60  th and 40 percentile, respectively, of the projected aggregate  diameters from TEM measurement. In general, the aggregate diameters from the TEM measurements were found to be larger than the mean mobility diameters measured from the nano-DMA. One of the possible explanations is that large particles that carry more than a single charge were removed by the impactor prior to entering the classifier in the SMPS system and hence, the DMA did not measure large particles that have diameters more than the upper size limit.  Table 3.9 Mobility Diameters and Aggregate Diameters for Cummins ISX Soot Mode I  DA from TEM Analysis (nm) Nano-DMA (rim)  414  43.1  38.6  Mode 2 40.9  43.8 (36.3)  36.5  Mode 3 44.4  41.2 (.38., 36.4  Table 3.10 summarizes the average primary particle diameters (dr) from the TEM measurement, and their 95% confidence interval. Similar as before, the confidence intervals were calculated from the t distribution. The primary particle diameters were measured in the range of 33 —34.1 urn across all engine loads.  65  Table 3.10 Mean Diameters of Primary Particles for Cummins ISX Soot Mode  RPM  Torque  Fuel Type  (N.m)  Natural Gas  1  1200  247  33.0±3.7  2  1200  207  33.8±4.5  3  1200  168.7  34.1±4.2  3.5.2 Fractal Properties of Soot Aggregates The samples of the fractal analysis results based on the projected aggregate dimensions for the ULSD and 1320 soot (without the oxidation catalyst) from the VW TDI engine are shown in Figures 3.19 and 3.20, respectively. These results were obtained using MATLAB program found in Appendix D. In general, the values of the fractal prefactor (kw) ranged from 1.30  —  1.39 with a standard deviation less than 0.3. These values were  in reasonable agreement with the estimates by Koylu et al. (1995).  66  ÷ 101  Model  1703 /d) 112 ___—N=l.30((LW)  *  / 0  Mode 2  —N= 1.35 *  Mode3  —N O  >  =  1.32 1t2 ((LW) .843 1 fd)  Mode4 N  =  1 / 2 d) .821  1.35 ((Ll  Mode 5  —N= 1.36 *  Made6 N  =  fd) 2 ‘ 1.33 ((LW) 11 101  Figure 3.19 Fractal Analysis Based on the Aggregate Dimensions for ULSD Soot without Oxidation Catalyst 7  1& +  Model /d)l .714 2 N = 1.30 ((LW)W  o  Mode2 N  =  1.35 ((LW) 11 /d) 2 ‘  *Mode3 /d) 1 ((LW) 1838 0  Mode4 —N = 1.35 ((LW) ’Fd) .833 1 0 ModeS 11 /d) .718 2 ‘ —N = 1.39 ((LW) *  Mode6 1.34 ((LW) /d 2 ’ 1  —N  10  )1.848 p  101  10°  1‘ (LW) Id 2  Figure 3.20 Fractal Analysis Based on the Aggregate Dimensions for B20 Soot without Oxidation Catalyst  67  The results of the fractal analysis based on the projected aggregate dimension method, including the fractal dimensions and their 95% confidence interval, for the soot from the VW engine are summarized in Table 3.11. These fractal dimensions were derived for the thirty-five aggregates from each sample. The confidence intervals were calculated from the least-square linear regression analysis. It should be noted that the uncertainty in the fractal dimensions obtained from the fractal analysis based on the projected aggregate methods can be mostly attributed to human error in estimating the number of the primary particles in the soot aggregates. The comparisons between the fractal dimensions of the ULSD and B20 soot were shown in Figure 3.21.  68  Table 3.11 Dr of TJTISD and B20 Soot Derived from Areate Dimension Method Mode  No Catalytic Converter  Torque (N.m)  Fuel Type ULSD  820  76.4 38.2 19.1  1.703±0.045 1.788 ± 0.050  1.714±0.042 1.773 ± 0.046  1.843  0.057  1.838 ± 0.053 1.833±0.056 1.718±0.067  1  3750  2  3750  3  3750  4  2600 1900  163.5 155  1.821±0.052 1.712±0.069  1900 3750 3750  37.7  1.848±0.058  76.4  1.698  0.054  1.847±0.055 1.705 ± 0.049  3750  38.2 19.1  1.774±0.052 1.842 ± 0.055  1.780±0.056 1.839 ± 0.058  5  2600 1900  163.5 155  1.826±0.051 1.720±0.049  1.822±0.054 1.722±0.060  6  1900  37,7  1.849±0.061  1.848±0.061  5 6 1 Cata’ytic Converter  RPM  2 3 4  ±  ±  Fractal Dimension (with 95% Confidence Interval) • IJLSD, No After Treatment . B20, No After Treatment 2.00  A ULSD, After Treatment  • B20, After Treatment  1.95  1.90  1.85  1.80  1  1.75  1.70  I  1.65  1.60 1 RPM Torque (Nm)  3750 76.4  2 3750 38.2  3  4  3750 19.1  2600 163.5  5 1900 155  6 1900 37.7  Engine Mode  Figure 3.21 Df of ULSD and B20 Soot Obtained from Aggregate Dimension Method  69  From Figure 3.21, both ULSD and B20 produced soot aggregates with identical fractal dimensions within experimental uncertainties. This finding supports the conclusion drawn by Koylu et al. (1995) that the DfjS relatively independent of the fuel type. Similarly, the presence of the oxidation catalyst did not seem to affect the fractal dimensions of the soot aggregates. Engine loads, however, appeared to have slight effects on the soot fractal dimensions. As the engine load was significantly increased (i.e., from engine mode 3 to mode 1 or from mode 6 to mode 5), the fractal dimension of the aggregates decreased. This means that the engine produced more stretched chain-like aggregates at high load. The explanation for the decrease of the fractal dimension with the engine load was proposed by Skillas et al. (1998). They suggested that aggregate growth is likely dominated by monomer-cluster aggregation through the ballistic mechanism at low load. At high load, the particle aggregates are likely to grow by the diffusion-limited mechanism where the cluster-cluster collision scheme starts to compete with the monomer-cluster aggregation, resulting in a decrease of the fractal dimensions.  Table 3.12 summarizes the fractal dimensions of the soot from the Cummins ISX engine  and their 95% confidence interval found using the projected aggregate dimension method. Figure 3.22 illustrates the fractal dimensions of the natural gas soot produced at different engine loads.  70  Table 3.12 Df of Natural Gas Soot Derived from Aggregate Dimension Method Mode  RPM  Torque (Nm)  Fuel Type Natural Gas  1  1200  247  1.722±0.057  2  1200  207  1.743±0.051  3  1200  168.7  1.776±0.054  Fractal Dimension (with 95% Confidence Interval) 1.0  1.85  1.8  1.75  U-  1.7  1.65  1.6 1  RPM Torque (N.m)  1200 247  2  1200 207  3  1200 168.7  Engine Mode  Figure 3.22 Df of Natural Gas Soot Obtained from Aggregate Dimension Method  As observed from Figure 3.22, the natural gas soot produced by the Cummins ISX engine also showed a decrease in the mean value of the fractal dimension as the load increased. The soot samples from Cummins ISX engine were collected under a limited range of engine loading conditions and hence, the variations in the fractal dimension with engine load were not as obvious compared to the soots produced by the VW engine.  71  In this study, the fractal analysis based on the pair correlation method (PCM) was also used to derive the fractal dimensions of the engine-emitted soot. Table 3.13 summarizes the mean values and 95% confidence intervals of the fractal dimensions obtained from the PCM for the soot from the VW TDI engine. The fractal dimensions listed in Table 3.13 are the average value of D from thirty-five aggregates. The confidence intervals are calculated from the t distribution.  Table 3.13 Df of ULSD and B20 Soot Derived from the PCM Mode  RPM  Torque  Fuel Type  (N.m)  ULSD  1  B2f.  1  3750  76.4  1.663 ± 0.040  1.670 ± 0.040  2  3750  38.2  1.733 ± 0.036  1.727 ± 0.037  No Catalytic  3  3750  19.1  1.784 ± 0.027  1.786 ± 0.028  Converter  4  2600  163.5  1.762 ± 0.037  1.757 ± 0.036  5  1900  155  1.686±0.040  1.695±0.042  6  1900  37.7  1.831 ± 0.028  1.827 ± 0.03  1  3750  76.4  1.656±0.041  1.705±0.041  2  3750  38.2  1.730 ± 0.037  1.731 ± 0.035  3  3750  19.1  1.783±0.025  1.780±0.033  4  2600  163.5  1.765±0.034  1.766±0.036  5  1900  155  1.688±0.042  1.691±0.040  6  1900  37.7  1.827±0.025  1.826±0.028  Catalytic Converter  From Table 3.13, the results of the fractal analysis based on the pair correlation method also showed that the fuel types (B20 and ULSD) and the use of the oxidation catalyst did not affect the fractal dimension of the soot. Moreover, it was also observed that Df decreased as the engine load increased.  72  Table 3.14 summarizes the results of the fractal analysis based on the pair correlation method for natural gas soot. As expected, a similar trend of the fractal dimension with respect to the engine load was observed.  Table 3.14 D of Natural Gas Soot Derived from PCM Mode  RPM  Torque  Fuel Type  (N.m)  Natural Gas  1  1200  247  1.722±0.057  2  1200  207  1.743±0.051  3  1200  168.7  1.776±0.054  Figures 3.23 and 3.24 illustrate the comparison between the fractal dimensions obtained from the projected aggregate dimension method and the pair correlation method for the ULSD soot (without oxidation catalyst) and the natural gas soot, respectively. The error bars in those figures represent the 95% confidence interval.  73  O (with 95% ConfIdence Interval) without Oxidation Catalyst  [.Roiected Aggregate Dinsion . Pair Correlation 2.001 1.95  1.90 1.85 1.80  a 1.75 U-  1.70  1.65 1.60  RR1 Torque (Nm)  1  2  3750 76.4  3750 38.2  3  4  3750 2600 19.1 163.5 Engine Mode  5  6  1900 155  1900 37.7  Figure 3.23 Comparisons between Df Derived from Aggregate Dimension Method and PCM for ULSD Soot without Oxidation Catalyst  D,for Natural Gas Soot (with 95% Confidence Interval)  Rojected Aggregate [rension. Pair Correlation 190  1.85  1.80  1.75  1.70  1 65  1.60  RPM Torque (Nm)  1  2  3  1200  1200  1200  207  168.7  247  Bigine Mode  Figure 3.24 Comparisons between Df Derived from Aggregate Dimension Method and PCM for B20 Soot without Oxidation Catalyst  74  Referring to Figures 3.23 and 3.24, it can be seen that the fractal dimensions derived from the projected aggregate dimension methods are in reasonable agreement with the fractal dimensions obtained from the pair correlation methods. In general, the mean values of the fractal dimensions obtained from the projected aggregate dimension method were roughly 3 % higher than those from the pair correlation method. It was expected that the fractal analysis based on the projected aggregate dimensions would produce higher Df since this method has tendency to overestimate the fractal dimension (Köylu et aL, 1995).  Finally, the fractal analysis results obtained in the present study are compared to the results of the previous attempts to measure the fractal dimension of engine-emitted soot in Table 3.15. The ranges of the fractal dimensions for the soot produced by the VW TDI and the heavy-duty Cunimins ISX engine are in a good agreement with the past TEM measurements.  Table 3.15 Comparison of D of Engine-Emitted Soot from TEM Measurements Study Present Study  Engine Type  Df  VW TDI engine, various loads and speeds  1.65  Cummins ISX engine, three operating conditions  1.68  Koylu & Neer (2006)  Six-cylinder diesel engine, various loads and speeds  1.7  Zhu et al. (2005)  Light-duty diesel engine, various loads and speeds  1.5  Weritzel et al. (2003)  Diesel engine, single operating condition  Lee et al. (2002)  Heavy-duty diesel engine, four operating conditions  1.85  —  1.78  —  1.9  —  1.7  —  1.7 1.8  —  1.9  75  4.0  4.1  DIESEL SOOT MICROSTRUCTURE  Introduction  The second objective of this study is to characterize the degree of graphitization present in the different soot. In order to reach this objective, Raman spectroscopy was employed in this research. The Raman spectroscopy analysis was a method of choice since it is sensitive to both the crystalline (long-range order) and molecular structures (short-range order) and hence, can be used to quantitatively distinguish between different types of soot based on their degree of graphitization.  In this study, the analysis of the Raman spectra of soot was performed with two different fitting procedures: two-band and five-band combination methods. The Raman spectral analysis based on the two-band combination (“G” and “D” bands at —4578 and —4340 ) has been routinely used for characterizing the soot microstructure (Rosen & 1 cm Novakov, 1978; Robertson, 1986; Dippel & Heintzenberg, 1999; Mernag, Cooney, & Johnson, 1984; Cuesta et al., 1994; Jahwari, Roid, & Casado, 1995; Williams et al., 2006). Some of these studies found that spectral analysis by curve fitting with two-band combination can be used to discriminate different type of soot according to their degree of graphitization (Dippel & Heintzenberg, 1999; Mernag, Cooney, & Johnson, 1984; Cuesta et al., 1994; Jahwari, Roid, & Casado, 1995; Zhu, et al., 2005; Williams et al., 2006). In particular, recent studies done by Williams et al (2006) and Zhu et a!. (2005)  76  employing the spectral analysis with two-band combination method successfully show that the degree of graphitization present in the diesel engine soot varies with the fuel type and engine load, respectively.  In contrast, the Raman spectral analysis based on five-band combination was recently introduced by Sadezky et al. in 2005. This method utilizes all reported first-order Raman bands of soot (G,D1,D2,D3,D4 at about 1580,1350, 1500, 1620 and 1200 cm-i respectively). Further investigations (Sadezky et al., 2005; Ivieva et al., 2007) show that the spectral parameters determined with the five band fitting procedure in particular the -  Dl band width  —  can be used to provide information on the structural order of several  different types of industrial carbon black such as Printex XE 2 and graphite bar. However, the spectral analysis using five-band combination has never been used to distinguish engine-emitted soot produced from different fuel types and engine operating conditions. Thus, this study is the first attempt to apply the Raman spectral analyses with the five-band combination method for investigating the degree of structural disorder of engine soot from different fuel types and engine loading conditions.  4.2  Experimental SectIon  4.2.1 Samples Soot Samples from Light-Duty Diesel Engine The engine soot was sampled from the 2001 four-cylinder Volkswagen 1 .9L turbocharged direct injection (TDI) engine. For the Raman spectroscopy study, the  77  particles were extracted from the diluted exhaust gas after the first stage of dilution. The particles were collected on 47 mm diameter teflon filter membranes with a 0.45 tm pore size (Apex Instrument TF1 121).  Only three different engine operating conditions were considered for the Raman spectroscopy study as listed in Table 4.1. Two different fuel types, ultra-low sulfur diesel (ULSD) with 15 ppm sulfur content and B20 blends (i.e., a blend of 20 wt% methyl esters in ultra-low sulfur diesel), were investigated. When the engine was run on the ULSD fuel, the engine tests were conducted with and without the presence of an AP Merit Catalytic Converter after-treatment system.  Table 4.1 VW TDI Engine Operating Conditions for Raman Study Mode  RPM  Torque (N.m)  1  3750  76.4  2  3750  38.2  3  3750  19.1  Soot Samples from Heavy-Duty Engine The soot samples were taken from a Cumniins ISX series heavy-duty six-cylinder, four stroke, direct-injection engine, modified to run on one cylinder. The engine was operated at three different operating conditions as listed in Table 4.2  78  Table 4.2 ISX Cummins Engine Operating Conditions for Raman Study Mode  RPM  Torque (N.m)  1  1200  247  2  1200  207  3  1200  168.7  As with the VW TDI engine tests, the particles were sampled from the diluted exhaust gas after the first stage of dilution. The particles were deposited on the teflon filter membranes (diameter: 47 mm, pore size: 0.45 tm) supplied by Apex Instrument.  Graphite For comparison purposes, the microstructure of the graphite was also analyzed with the Raman spectroscopy in this study. The graphite sample was available in the form of  highly oriented pyrolytic graphite (HOPG, grade ZYC, from Mikromash). The properties of the HOPG used in this study are tabulated in Table 4.3. It should be noted that the term “mosaic spread”, which originates from X-ray crystallography, represents a measure of highly ordered the HOPG is: the lower the mosaic spread, the more highly ordered the HOPG (Mikromash, 2008).  Table 4.3 ISX Properties of HOPG  79  Graphite Aerosol Generator (GfG-1000) Soot Artificial soot generated by a Palas GfG-1000 graphite aerosol generator was also included in the Raman spectroscopy analysis for comparison. The GfG 1000 aerosol generator produces the soot from pure graphite. During the operation, the aerosol generator applies high voltage to generate sparks between two graphite-coated electrodes and subsequently, vaporize a small amount of graphite. The resulting vaporized graphite is then transported by a constant stream of argon to the outlet of the instrument. During the transportation, the vaporized graphite re-condenses back into solid carbon, creating very fine particles of pure carbon. These fme particles will eventually coagulate to form agglomerates during further transport. The soot concentration in the output of aerosol generator is controlled by adjusting the argon flow rate and the spark frequency. Detailed information on the Graphite Aerosol Generator can be found in the operating manual provided by the manufacturer (Palas, 2003).  4.2.2 Raman Spectroscopy The acquisition of Raman spectra were performed using Renishaw inVia Raman microscope system. The inVia Raman system incorporates a Leica light microscope with three objectives (5x, 20x, and 50x magnification). The microscope optics are mainly used to focus the incident laser beam onto to the sample and collect the backscattering light. The Raman microscope employed in this study is equipped with a Helium-Neon (He-Ne)  laser (20 MW at 633 nm) as the primary excitation laser, a holographic Notch filter and a dual-grating spectrograph with RenCam CCD detector (578 x 385 pixel).  80  The Raman spectral intensity and wavelength were calibrated with the silicon wafer by utilizing the first order Stokes Raman of pure Si at 520 cm-i as shown in Figure 4.1. The calibration is achieved when the peak of Si Spectra reaches at least 4000 counts.  Rarnan Spectra of Silicon 6000  5000  4000  0 C.)  3000  2000  1000  0  -  200  —‘—r—’---—-————---  -  300  400  500  -  600  700  800  Raman Shift (cm ) 1  Figure 4.1 Calibration Raman Spectra of Silicon The spectra of the sample were recorded in the range of 500  —  2000 cm’ (Raman shift)  with a spectral resolution of—4.5 cm . in order to find the optimum measurement 1 condition, the power of the excitation laser (1 to 100% relative intensity), spot diameter (0 to 100% defocusing) and exposure time were varied. Since the soot samples are quite susceptible to thermal degradation, they cannot be exposed to incident laser continuously. Hence, the Raman spectrometer was mostly operated in non-continuous scanning mode to prevent the laser from damaging or burning the samples. From the optimization, the highest quality and reproducibility of the soot spectra were typically achieved with a fully  81  focused laser beam (diameter of laser spot  1-2 urn) at 10% power and an exposure time  of4x30s.  The preliminary Raman measurement done on the soot sample revealed the presence of contaminant peaks in the Raman spectrum over the Raman shift range of 500  —  2000 cm’  as illustrated in Figure 4.2. In order to identify these extra peaks so that they could be excluded from the Raman spectral analysis, the Raman spectroscopy was calibrated against a pure gold sample utilizing the fact that gold should not exhibit any first-order Raman bands over the range of interest. The detailed Raman spectroscopy calibration using the gold sample can be found in Appendix E.  —Soot  —  5000  Gold  -  4500  Contaminant Peaks  g2500 C.)  //  A  •  2000 1500  1000  50: 500  700  900  1100  1300  Raman Shift  1500  1700  1900  (1)  Figure 4.2 Contaminant Peaks in the Recorded Raman Spectra  82  The Renishaw WIRE 2.0 software running under GRAMS/32 (Galactic, Levenberg— Marquardt non-linear least-squares fitting algorithm) was used to control the Raman system and perform the spectral analysis. The spectral parameter of the soot was determined by curve fitting after the baseline correction with the software program GRAMS/32. For consistency, a linear baseline correction which connects the average of the first 25 data points in the spectra to the average of the last 25 was applied for all recorded spectra. To ensure the reproducibility of the curve fit, the fitting procedure was repeated five times for each Raman spectrum. To avoid sample bias, ten spectra were recorded at different positions on the Teflon filter for each sample. The mean value and standard deviation of the spectral parameters presented in this study are derived from those spectra.  4.3  Raman Spectra of Soot and Graphite  Figure 4.3 shows the typical characteristic of the first-order Raman spectra of the investigated soot and graphite samples. The spectra are offset to make the comparison clearer in Figure 4.3.  83  First-Order Raman Spectra of Carbonaceous Materials 12000  10000  8000  C  Natural Gas  6000  ULSD, After Treatment 4000  ULSD, No After Tieatment B20, No After Treatment  2000  GfG Soot Graphite 500  1000  1500  2000  Raman Shift (c ) 1  Figure 4.3 First-Order Raman Spectra of Different Carbonaceous Materials  The Raman spectrum of the HOPG usually exhibits only one first-order band, G (“Graphite”) band at around 1580 cm’ representing the characteristics of highly ordered graphite with undisturbed graphite lattice. However, an additional band (D or “Defect” band) is also observed in the spectrum of the HOPG investigated in this study. The presence of the D band, which is known to be a characteristic for disordered graphite, is likely due to the high disorientation of the graphene sheets present in the HOPG grade ZYC. It should be also noted that the typical graphite spectrum does not exhibit peaks at -  1000 cm 1 and  -  1250 cm’. Those extra peaks in Figure 4.3 are the aforementioned  contaminant peaks.  84  For other carbonaceous particles such as soot which can have highly disordered graphitic structures, the first-order Raman spectra is typically comprised of two broad and strongly overlapping bands that peak at  1350 cm’ (D or “Defect” peak) and  1580 cm (0 or  “Graphite” peak) as illustrated in Figure 4.3. The 0 (“Graphite”) band at 1580 cm-I corresponds to an ideal graphitic lattice (E2g symmetry) stretching vibrations, and is a  typical characteristic of a natural graphitic crystal (Tuinstra & Konig, 1970; Ferrari & Robertson. 2000: Sadezky et aL. 2005). The D (“Defect”) band at  1350 cm 1 is usually  used to characterize the disordered graphitic lattices and is occasionally associated with polyaromatic ring vibrations (Popovitcheva et al., 2002). The intesity of the D band is known to grow relative to the G band with increasing degrees of structural disorder in carbonaceous particles (Robertson, 1986; Dippel & Heintzenberg, 1999; Sadezky et al, 2005).  Further experimental observations and theoretical calculations revealed that the peak at 1 actually consists of Dl and D4 bands. The Dl band (at ‘-1350 cm’) is the most 1350 cm intensive among the defect bands and has been associated with a vibration mode involving graphene layer edges with Aig symmetry crystal (Tuinstra & Konig, 1970; Ferrari & Robertson, 2000; Sadezky et al., 2005). The study by Wang, Alsmeyer and MeCreery (1990) has also suggested that the Dl band likely originates from carbon atoms at the edge of graphene layers for polycrystalline carbonaceous materials containing large numbers of small graphitic crystallites.  85  The D4 band usually appears at  1180 cm 1 as the shoulder of DI band. D4 is the  weakest band compared to other defect bands, and usually corresponds to the vibrations  of disordered graphite lattices (Aig symmetry), sp -and sp 2 -hybridized carbon bonds, C—C 3 and C=C stretching vibrations of polyenes, and ionic impurities (Cuesta et al., 1994; Dippel et al., 1999; Sadezky et a!., 2005; Ivieva et a!., 2007).  Further investigations by Cuesta et al (1994), Jahwari et al. (1995) and Sze et al. (2001) propose that the peak at 1580 cm’ consist of not only the G band, but also the D2 band which can be observed at 1620 cm as a shoulder on the G band. The D2 band is usually attributed to the lattice vibration analogous to that of the G band but involving surface graphene layers with E2g symmetry (Ferrari & Robertson, 2000; Sadezky et al., 2005; Ivleva et al., 2007).  Several studies also suggest that the high signal intensities between the two main peaks of the Raman spectra of carbonaceous particles can be attributed to another band at 1500 cm. This band is usually referred to as D3 band and can be attributed to the amorphous carbon content of soot including organic molecules, fragments, and functional groups (Cuesta et al., 1994; Jawhari et al., 1995; Dippel et al., 1999; Sadezky et al., 2005; Ivieva et al., 2007).  4.4  Raman Spectra Analysis  In order to determine the spectral parameters of the investigated soot, a quantitative analysis was performed by curve fitting with two different band combinations: two-band  86  and five-band combinations. The detailed procedure for the curve fitting using both methods will be discussed below  4.4.1 Two-Band Combination Method Curve fitting with the two-band combination is the most commonly used method for analyzing the Raman spectra of carbonaceous particles (Memag, Cooney, & Johnson, 1984; Zhu et al., 2005; Williams et al, 2006). This method offers a quick and robust analysis of the degree of graphitization present in the sample. For spectral analysis with the two-band combination, the spectra were fitted with a combination of “G” and “D” band at 1578 and —1340 cm’ respectively as shown in Fiaure 4.4. The “G” band was not separated from the D2 and D3 bands while the “D” band consists of Dl, D3 and D4 bands. Hence, the intensity ratio serves as indicator for the degree of graphitization since past research has proved that the “D”/”G” band intensity ratio is a measure of the ordering of the sp 2 phase present in the carbonaceous materials (Chhowalla et al., 2000).  87  Two-Band Combination Method 2500  2000  1500  1000  500  0 500  600  700  800  900  1000  1100  1200  1300  1400  1500  1600 1700  1800  1900 2000  Raman Shift (cm ) 1  Figure 4.4 Exemplary Curve Fit for Raman Spectra Using Two-Band Combinations  Despite the fact that the two-band combination method has been routinely used for spectral analysis, the line shapes of both “D” and “G” bands have never been clearly reported. Therefore, all possible band combinations were tested in this study to determine which band combination would provide the best fit. Table 4.4 lists the different band combinations tested.  88  Table 4.4 Band Combinations Tested for Curve Fitting with Two-Band Combination (Line shape: G = Gaussian, L = Lorentzian) Band  Initial Position (cm -1)  (I)  (II)  (III)  (IV)  “G”  1578  G  L  G  L  “D”  1340  G  [L  L  G  The chi-square goodness-of-fit tests were performed using the GRAM/32 which utilizes a curve fitting algorithm based on the Levenberg—Marquardt. The goodness-of-fit is represented by the reduced x 2 value. For an ideal fit between the spectrum and the curve fit, the x 2 is between 1 and 3, the curve fit converges 2 value is equal to unity. If the x toward the spectrum. The  2 value larger than 3 implies that the curve fit fails to converge x  toward the spectrum at the end of the iteration process.  The goodness-of-fit tests for the band combinations were conducted for nine different soot samples taken from both the VW TDI and ISX Cummins engine. Table 4.5 summarizes the results of the goodness-of-fit tests. The  listed in Table 4.5 represent the  average values over 10 spectra.  89  Table 4.5  Values for the Raman Spectra Fitted with Two-Band Combination Samples  (I)  (II)  (III)  (IV)  ULSD, Mode 1  2.84  9.14  6.64  2.22  ULSD, Mode 2  2.74  7.67  4.87  1.94  ULSD, Mode 3  3.44  9.38  6.07  2.34  B20, Mode 1  2.86  6.77  5.83  1.96  B20,Mode2  3.12  8.51  6.19  2.32  B20, Mode 3  2.95  7.06  4.86  2.25  Mode 1  2.81  7.32  4.51  2.16  Mode 2  2.89  6.57  5.88  1.99  Mode 3  2.98  8.83  6.02  2.27  [  VW TDI  ISX Cummins  The curve fitting based on the two-band combination (“G” and “D” band) typically is not able to produce an ideal fit (i.e.,  is close to unity) due to the fact that the fit will not  properly converge toward the spectrum at 1400 cm to --‘1 500 cm’. For the four different band combinations tested, only band combination (I) and (IV) yielded an average value of lower than 3 (the minimum but no convergence of the Levenberg— Marquardt fit). The lowest value of 2 is clearly achieved with combination (IV), which assumes a Lorentzian shape for “G” and a Gaussian shape for “D” bands.  44.2 Five-band Combination Method Previous Raman spectral analysis of soot usually considered only three bands (G, D, and either D2 or D3). As a result, the analysis often produced a curve fit which failed to converge toward the spectrum. Recently, Sadezky et al (2005) tested nine different band 90  combinations and found that the best fit to the Raman spectra of soot and related carbonaceous materials was obtained by curve fitting with five bands (G, Dl, D2, D3, and D4) as illustrated in Figure 4.5. It was reported that the Dl band width allows tbr the discrimination between different types of soot.  Five-Band Combination ivethou 2500 Dl  2000  G 1500  1000  D3  D4  D2 500  0 500  600  700  800  900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Raman Shift (cn  Figure 4.5 Exemplary Curve Fit for Raman Spectra Using Five-Band Combinations  The analysis of spectral parameters by curve fitting with the five band combination, however, has never been used to investigate engine soot produced from different fuel types and engine operating conditions. Hence, a goodness-of-fit test was also conducted to determine the band combination that would provide the best fit. In addition, this test was also purposely used to investigate the shape of the D3 band. The majority of the previous studies (Dippel & Heintzenberg, 1999; Cuesta et al., 1994) assumed that the D3 band has a Lorentzian shape. However, Jawhari and his coworkers (1995) proposed a  91  Gaussian line shape for the D3 band because that the amorphous carbon in interstitial places in the disturbed graphitic lattice of soot exhibited a Gaussian distribution. Curve  fitting with eight different band combinations of first-order Raman bands was tested in this study. Table 4.6 lists the tested band combinations with the applied line shapes and their initial positions.  Table 4.6 Band Combinations Tested for Curve Fitting with Five-Band Combination (Line shape: G = Gaussian, L = Lorentzian) Band  Initial Position (cm -1)  (I)  G  1578  L  Dl  1350  fl”  (II)  (VI)  (VII)  (VIII)  G  L  G  L  L  G  L  G  L  T  (1  T  fi  T  -  -  L  G  G  L  G  L  G  L  (III)  (IV)  L  G  L  L  L  G  16’fl  -  -.  -  _D3  1500  L  G  G  1)4  1180  [L  L  G  (V)  Table 4.7 summarizes the results from the goodness-of-fit tests. The goodness-of-fit obtained with different band combinations is indicated by the reduced x 2 value. For each sample, ten spectra were analyzed during the test.  92  Table 4.7 x 2 Values for the Raman Spectra Fitted with Two-Band Combination (I)  (II)  (III)  (IV)  (V)  (VI)  (VII)  (VIII)  ULSD, Model  2.11  1.94  1.96  7.45  2.82  1.43  1.63  1.21  ULSD,Mode2  2.25  1.83  1.87  6.39  2.87  1.62  1.71  1.13  ULSD, Mode 3  1.98  2.05  1.99  7.12  2.78  1.46  1.56  1.26  B20, Mode 1  1.95  2.07  2.05  6.83  2.82  1.57  1.68  1.16  B20,Mode2  2.17  2.01  2.01  7.01  2.95  1.59  1.54  1.18  B20, Mode 3  1.95  2.06  2.06  7.04  2.91  1.63  1.65  1.22  ISX  Mode 1  2.01  1.62  2.07  6.86  2.83  1.56  1.52  1.14  Cummins  Mode 2  1.98  1.84  2.08  6.95  2,87  1.59  1.58  1.17  Mode 3  2.08  1.67  1.91  6.27  2.97  1.64  1.69  1.31  Samples VWTDI  Averaging over all the investigated spectra yields 2.1 ±  ±  0.1 for band combination (I), 1.9  0.2 for band combination (II), 2.0 ± 0.4 for band combination (III), 2.9 ± 0.3 for band  combination (V), 1.6± 0.2 for band combination (VI), 1.6±0.1 for band combination (VII), and 1.2 ± 0.1 for band combination (VIII). Only combination (IV) produced an average  value higher than 3.  This result implies that the soot spectra are best fitted with the combination of all reported first-order Raman bands (G, Dl, D2, D3, D4 at about 1580, 1350, 1500, 1620 and 1200 cm’). Moreover, in concurrence with the finding by Sadezky et al. (2007), the shape of the D3 band indeed is Gaussian; while the other four bands (G, Dl, D2, and D4) are best fitted with Lorentzian-shaped curves.  93  4.5  Results and Discussion  For all investigated soot samples and carbonaceous materials, the spectral parameters  were obtained by curve fitting with the two-band (Lorentzian-shaped “G” and Gaussianshaped “D” bands) and five-band (Lorentzian-shaped G, Dl, D2, D4 bands and Gaussian-shaped D3 bands). The detailed data with mean values and standard deviation of the band positions (Stokes Raman shift), full width at half maximum (FWI-IM), and intensity (peak area) ratios of the spectral parameters determined using the five-band combination can be found in Appendix F; while the important spectral parameters determined by curve fitting with the two-band combination will be outlined below.  Table 4.8 summarizes the mean values and standard deviations of the “G” band position (Stokes Raman shift), “D” band positions, and “D”/”G” band intensity ratios (I”jy’/I””) obtained from the two-band combination method for the soot collected from the VW TDI engine. The spectral parameters of the soot produced by the aerosol generator (GFG 1000), and highly oriented pyrolytic graphite (HOPG) are also listed in Table 4.6 as references. Statistically, no significant differences are observed for the “G” and “D” band positions between the biodiesel soot, the GfG-1000 soot, and HOPG. As expected, the “D”I”G” band intensity (peak area) ratio of the HOPG was significantly lowered compared to the biodiesel soot and highly amorphous spark discharged GfG-1000 soot indicating higher structural order.  94  Table 4.8 “D”/”G” Ratio for Soot from VW Engine Sample  Mode  “D” Band  “G” Band  Location  Location  ULSD  1  1342.3±4.3  1587.9±4.8  3.18*0.07  No After Treatment  2  1341.6 ± 3.8  1588.4 ± 4.4  3.61 ± 0.08  3  1341.5±5.3  1586.4±4.9  4.45±0.10  B20  1  1340.2 ± 6.3  1585.2 ± 5.4  3.68 ± 0.27  No After Treatment  2  1340.8 ± 4.5  1586.1 ± 4.6  4.44 ± 0.28  3  1340.6 ± 2.7  1588.9 ± 4.0  5.15 ± 0.32  ULSD  1  1341.2 ± 5.2  1587.2 ± 3.7  3.19 * 0.14  AfterTreatment  2  1343.0±4.8  1587.2 ±4.2  3.70±0.15  3  1341.8 ± 5.1  1587.2 ± 3.9  4.51 ± 0.19  HOPG  1341.4 ± 1.6  1581.9 ± 1.8  0.37 ± 0.01  GfG-1 000  1342.5 ± 3.1  1580.6 ± 8.7  4.28 ± 0.20  Table 4.9 lists the width of the Dl band obtained from the curve fitting with the five-band combination for the soot from the VW TDI engine and other investigated carbonaceous materials. The Dl band FWHM of HOPG is significantly lowered compared to the diesel soot, which is consistent with the result obtained from the two-band combination showing a higher degree of graphitization present in the HOPG.  95  Table 4.9 Dl Band FWHM for Soot from VW Engine Sample  Mode  DI Band FWHM  ULSD  1  180.05±13.72  No After Treatment  2  229.07 ± 11.02  3  275.05±15.28  B20  1  279.52± 32.56  No After Treatment  2  357.55 ± 30.26  3  440.05±34.31  ULSD  1  185.8±20.45  After Treatment  2  232.27 ± 22.00  3  288.62±21.41  HOPG GfG-1 000  37.74 ± 8.52 327.92 ± 27.53  The comparisons between the results of the spectral analysis obtained with the two-band combination and five-band combination are illustrated in Figure 4.6 and 4.7 respectively. It was confirmed that the analysis results based on the two-band combination are consistent with the results based on the five-band combination.  96  I,II< (with Std. Deviation) • LLSD, No Ox. Catalyst A ILSD, Ox. Catalyst x GFG Soot  • B20, No Ox Catalyst A Graphite  6  RPv1 Torque (Nm  Mode 1 3750 76,4  Mode 2 3750 38.2  Mode 3 3750 19  Graphite  GFG Soot  Sample  Figure 4.6 “D”I”G” Baud Intensity Ratios for Soot from VW Engine  Dl FWHM (with Std. Deviation) • B20, No Ox. Catalyst • ULSI), No Ox. Catalyst A ULSD, Ox. Catalyst A Graphite x GFG Soot 500 450 400  f  350 300  -  250 200 150 100  RPM Torque (N1m  Mode 1 3750 76.4  Mode 2 3750 38.2  Mode 3 3750 19.1  Graphite  GFG Soot  Sample  Figure 4.7 Dl Band FWHM for Soot from VW Engine  97  From both the two-band and five-band combination method, it was found that the biodiesel (B20) fuel produced soot with higher structural disorder (i.e., lower degree of graphitization)  —  indicated by increase in the “D”/”G” band intensity (peak area) ratios  and Dl band FWHM— compared to the ULSD soot. This result is in agreement with the previous Raman analysis showing greater disorder for biodiesel soot (Williams et al., 2006). In addition, this finding also supports the HRTEM observation that the soot produced from biodiesel has a less-ordered structure (Vander Wal & Tomasek, 2004; Boehman, Song, & Alam, 2005).  Boehman, Song and Alam (2005) proposed preliminary hypothesis to explain the presence of a more-amorphous structure in the biodiesel soot. Their first proposed explanation is that the fuel composition are likely responsible for the amorphous structure in the B20 soot. In terms of composition, biocliesel fuels usually have lower polycyclic aromatic hydrocarbons (PAH) contents which results in less PAH present during the soot inception process. Based on the study by Hurt and his coworkers (2000), the transition of the soot microstructure into the typical ordered shell/core arrangement can only happen when the threshold molecular weight requirement is achieved. For biodiesel soot, the lack of PAH available during the soot inception period can possibly lead to a failure to meet this threshold requirement and prevent the disorder/order transition of the soot microstructure.  The other explanation for the difference in the microstructure between B20 and ULSD soot is related to the differences in the soot growth mechanism caused by different fuel  98  decomposition chemistry. A previous study done by Vander Wal and Tomasek (2003) has proven that the differences in soot mierostructure can be attributed to the different growth species formed during the fuel decomposition process. A soot growth process dominated by high-molecular weight PAR deposition likely leads to a more amorphous structure. The reason is because the coalescence process between various PAR species likely create atomically incongruous surfaces. During the growth process, this coalescence processes will tend to occur continuously, resulting in a structurally disordered solid (Vander Wal & Tomasek, 2003). On the other hand, a soot growth process dominated by smaller species such as light acetylene (C ) leads to a graphitic H 2 structure (Vander Wa! & Tomasek, 2003). This is because the growth of smaller species like C 2 tends to facilitate the formation of more-ordered graphene layers through the I{ACA (hydrogen abstraction carbon addition) mechanism. It is suspected that the soot produced by biodiesel likely favors the coalescence growth mechanism, leading to a more amorphous soot structure.  The Raman analysis using both the two-band and five-band combination methods also suggested that a reduction in engine load led to a decrease in the degree of graphitic structural order in the soot samples as illustrated in Figure 4.6 and 4.7. This finding is consistent with the previous results of the Raman analysis by Zhu et al (2002), which found that the FWHM D increases as the engine load decreased indicating higher disorder of graphitic structure. Moreover, the Raman analysis results presented here offers additional evidence, supporting a HRTEM observation by Lee et al. (2002) that the soot sampled at the high engine loads exhibited the graphitic structures while the soot  99  samples from the lower engine loads displayed mostly amorphous structures. An increase in the order of graphitic structures with engine load is likely due to the fact that diesel particulates become more stable and organized in internal structure at high engine loads.  From both Figure 4.6 and 4.7, it was also observed that the use of biodiesel (B20) fuel as well as the oxidation catalyst resulted in noticeable increases of the standard deviations of the “D”/”G” ratio and Dl FWFIM. The increased standard deviations signified that much greater variations were present in the soot domain structure.  The mean values and standard deviations of the “D”/”G” band intensity ratios and Dl  FWHM obtained from the Raman analysis with two-band and five-band combinations for the natural gas soot produced by the Cummins ISX engine are summarized in Table 4,10. The mean values of the “D” band and “0” band positions (Stokes Raman shift) observed for natural gas soot ranged from 1334.2 cm 1 to 1347 cm 1 with standard deviations (s.d.) up to 6.1 cm . The mean values and standard deviations of the spectral parameters 1 obtained with the five-band combination method are listed in Appendix F.  Table 4.10 “D”I”G” Ratio and Dl Band FWHM for Soot from Cummins ISX Engine Samp’e  Natural Gas  Mode  G’ 4 ‘ 1  Dl Band FWHM  1  3.29 ± 0.20  218.26 ± 21.3.4  2  3.84 ± 0.22  282.91 ± 22.49  3  4.40 ± 0.20  348.97 ± 22.37  100  For comparison purposes, Figure 4.8 and 4.9 show the results of the spectral analysis obtained with the two-band combination and the five-band combination, respectively. From both Figure 4.8 and 4.9, it was verified that the results from the two-band combination method are in agreement with the results from the five-band combination method.  (with Std. Deviation) for Natural Gas Soot  ‘D”G  5 4.5 4  3 2.5  1.5  0.5 0  RPTVI Tórque(N.m)  Fbde  Mode 2  Mode 3  1200  1200 207  1200 168.7  247 Sample  Figure 4.8 “D”I”G” Band Intensity Ratios for Soot from Cummins ISX Engine  101  r  01 PWHM (with Std. Deviation) tor Natural s Soot 400 350 300 250 200 150  RR Torque(N.m)  Mode 1  Mode 2  Mode 3  1200  1200 207  1200 168.7  247  I  Sam pie  Figure 4.9 Dl Band FWHM for Soot from Cummins ISX Engine Similar to the trend observed for B20 and ULSD soot, the disorder of graphitic structures present in the natural gas soot also decreased with engine load. These findings provide further confirmation that increased engine load caused increases in the degree of the graphitization of the soot.  An earlier study by Sze et al. (2001) has reported that the changes in the Raman shift of Dl bands can be attributed to structural differences. Gruber, Waldeck-Zerda, and Gerspacher (1994) have also suggested that a decrease of the Raman shift of the “G” band (comprising both G and D2 bands) indicates an increase in the degree of graphitization of the material. However, these trends were not observed in this study. For example, no significant differences could be observed between the Dl and “G” band  102  positions of the highly amorphous spark discharged (GfU- 1000) soot and more graphitic ULSD soot.  Some previous studies (Cuesta et al., 1994; Jawhari, Roid, & Casado, 1995; Ivieva et a!., 2007) have also proposed that a decrease of ID3/’G indicates an increase in graphitic structural order of the sample. This correlation, however, could not be observed and verified in this study. For instance, there were generally no significant differences observed between the ID3”G ratios of the more graphitic ULSD soot and the more amorphous B20 soot as shown in Figure 4.10.  ‘D3”G  (with Std. iJeviazio 11  ULSD, No Ox. Catalyst • 820, No Ox. Catai] 0.7  1  0.6 0.5 0.4 CD  —  0.3 0.2  0  Engine Mode  Figure 4.10 D3JG Band Intensity Ratios for Soot from VW Engine  103  o  CONCLUSIONS AND RECOMMENDATIONS  The morphology and microstructure of the engine-emitted particulates were investigated in this research. The study was conducted on two different engines: a 1 .9L Volkswagen Turbo Direct Injection (TDI) engine and a Cummins ISX heavy-duty engine using the Westport pilot-ignited direct-injection natural-gas fuelling system. The Volkswagen TDI  engine was operated with and without a catalytic converter, using two different fuel types (ULSD and B20) and six speed/load combinations; while the Cummins ISX engine was tested at three different speed/load combinations. The soot samples from both engines were collected on TEM grids and Teflon filters for the transmission electron microscopy (TEM) and Raman spectroscopy studies, respectively.  Transmission electron microscopy (TEM) was used to image diesel soot aggregates and derive the fractal properties of the aggregate. A Matlab-based image processor was developed to extract the morphological information of the soot aggregate from the TEM image. The projected aggregate diameters measured from the TEM were compared to the geometric mean mobility diameters of the soot aggregates measured by the SM?S. It was found that the aggregate diameters from TEM measurements were consistent with the SMPS results. For the soot collected from VW engine, both measurements suggested that the presence of an oxidation catalyst and biodiesel blend tends to decrease the mean diameter of the aggregate.  104  Two different methods of fractal analysis were applied to estimate the fractal dimensions of the soot aggregates: the projected aggregate dimension method and the pair correlation method. For the soot collected from VW engine, the results of the fractal analysis from both methods showed that: 1. The fuel types (ULSD and B20) and the use of an oxidation catalyst do not affect the fractal dimensions of the soot aggregates 2. The engine load has a slight influence on the fractal dimension (Df). The fractal dimensions of the soot started to decrease as the engine load was significantly increased. For the natural gas soot from the Cunimins ISX engine, the results of the fractal analysis based on both the projected aggregate dimensions and the pair correlation method also showed a decrease in the mean value of the fractal dimension as the load increased.  Microstructural characterization of the engine-emitted soot was investigated by means of Raman spectroscopy. Raman spectroscopy was applied to quantify the degree of structural disorder present in the soot. The spectral parameters of different soot samples were determined by two different methods: two-band (“G” and “D” at —4578 and —1 340 cm’) and five-band (G, Dl, D2, D3, D4 at about 1580,1350, 1500, 1620 and 1200 cm’ respectively) combinations. For the spectral analysis with two-band combination, it was determined that a combination of a Lorentzian-sbaped “G” band and a Gaussian-shaped “D” band was best-suited for the first-order spectra of soot; while a combination of provide the best fit for the five-band combination method.  105  For the soot collected from VW engine, the results obtained from the two-band and fiveband combinations showed that: 1. B20 soot exhibits greater structural disorder (i.e., a lower degree of graphitization) 2. The degree of graphitization of the soot increases with increasing engine load. Likewise, the Raman analysis of the natural gas soot from the Cummins ISX engine also showed that the degree of the structural order of the soot increased with engine load.  The main objectives set for this study have been achieved. The morphological properties of the engine-emitted soot were investigated by means of TEM analysis. The fractal analysis using the projected aggregate dimension and pair correlation methods were successfully applied to estimate the fractal properties of the soot. The present study also demonstrated the applicability of Raman spectroscopy for semi-quantitative analysis of soot microstructures, particularly to discriminate different types of soot based on their degree of graphitization. In addition, this study provided quantitative evidence that a combination of all five reported first-order Raman bands of soot (Lorentzian-shaped G, Dl, D2, and D4 bands with Gaussian-shaped D3 band) was best-suited for the first-order spectra and hence, has to be taken into account for a complete interpretation and analysis of first-order Raman spectra of the soot produced by the engines.  The information on the morphological properties and microstructure obtained in the present study should also be useful for further assessing the optical properties of the soot aggregates. The fractal properties found in this study can be applied to the Rayleigh Debye-Gans (RDG) theory for predicting both the scattering and absorption for the  106  engine-emitted soot aggregates. The review on the application of the RDG scattering approximation can be further found in a paper by Sorensen (2001). Likewise, the spectral parameters from the Raman analysis can also be used to estimate the absorptivity of the soot aggregates. The study by Chhowalla et al. (2000) provides an excellent investigation on how Raman DIG band intensity ratio varies inversely with the square optical gap (defined as the difference between the highest-energy occupied molecular orbital and the lowest-energy unoccupied molecular orbital present in the soot). In turn, the absorption coefficient of the soot can then be directly calculated from the optical gap. It should be noted that in most cases, the changes in the aggregate scattering cross section and absorption coefficient caused by the variations of the fractal dimension and DIG ratio, respectively, are relatively small (within 1%). Bond and Bergstrom’s 2006 review on the light absorption of carbonaceous particles should be consulted for further details on the calculation of the optical properties of soot aggregates.  Based on the experimental observations, the following recommendations for future researches related to the characterization of carbonaceous particulates morphology and microstructure of are given: 1. It is recommended that additional soot samples from the natural gas engine be collected under a wider range of operating conditions, In this study, most of the soot samples from Cummins ISX engine were collected under limited range of engine loads. Hence, the influence of engine load on the fractal dimension was not that obvious. By analyzing more samples from the natural gas engine under a wider range of operating conditions, the effect of engine load on the fractal  107  dimension can be more clearly determined.  2. Other method for analyzing the degree of graphitization present in the soot should also be applied to independently verify the Raman spectroscopy results One of the recommended approaches is to apply X-ray diffraction on the soot samples. Similar to Raman spectroscopy, the X-ray diffractograms can provide fmgerprint diffraction pattern for different types of soot. The distinctive reflections from the diffraction pattern combined with the Debye-Scherrer formula can be used to estimate the average size of the graphite-like crystalline domains present in the sample and indicate the degree of graphitization of the soot. Other commonly used method is to employ the electron energy loss spectroscopy (EELS) for characterizing the hybrid electronic configurations such 2 or sp as sp 3 contained in the soot. The discrimination of the degree of graphitization of soot can be done by analyzing the carbon K-edge spectra. In addition, it is also recommended that the Raman spectroscopy analysis should also be performed on additional commonly used reference materials such as Printex XE2 and Monarch 120. This will allow the Raman analysis results from the present study to be compared with the past Raman measurements.  3. Lastly, future experiments should consider collecting soot produced by an engine that can run on ULSD, B20 and natural gas fuel so that the influence of the natural gas on both soot morphology and microstructure can be investigated.  108  REFERENCES  Ackerman, S., Toon, 0. B., Stevens, D. E., Heymsfield, A. J., Ramanathan, V., & Welton, E. J. (2000). Reduction of tropical cloudiness. Science, 288, 1042  —  1047.  Albrecht, B. A. (1989). Aerosols clouds and microphysics. Science, 245, 1227—1230.  Andrea, M. 0., Rosenfeld, D., Artaxo, P., Costa, A. A., Frank, G. P., Longo, K. M., et a!. (2004). Smoking rain clouds over the Amazon. Science, 303, 1337  —  1342.  Barone, A.C., D’Alessio, A., & DAnna, A. (2003). Morphological characterization of the early process of soot formation by atomic force microscopy. Combustion and Flame, 132, 181  -  187.  Binnig, G., & Rohrer, H. (1985). Scanning tunneling microscope. Scientific American, 253, 50-56.  Biscoe, J., & Warren, B. E. (1942). An X-ray study of carbon black. Applied Physics. 13, 364—371.  Blanquart, G., & Pitscb, H. (2007). Thermochemical properties of polycyclic aromatic hydrocarbons (PAH) from G3MP2B3 calculations. Physics and Chemistry, 111, 65106520.  Boehman, AL, Song, J., & Alam, M. Impact of biodiesel blending on biodiesel soot and the regeneration of particulate filters. Energy Fuels, 1857— 1864.  109  Bond, T.C., & Bergstrom R.W. (2006). Light absorption by carbonaceous particles: an investigative review. Aerosol Science and Technology, 40, 27—67.  Boucher, 0., & Lohmann, U. (1995). The sulphate-CCN-cloud albedo effect: a sensitivity study with two general circulation models. Tellus, 47, 281  —  300.  Bouzerar, R., Amory, C., Zeinert, A., Benlahsen, M., Racine, B., Durand-Drouhin, 0., & Clin, M. (2001). Optical properties of amorphous hydrogenated carbon thin films. NonCrystalline Solids, 281, 171  —  180.  Calcote, H. (1981). Mechanisms of soot nucleation in flames  —  a critical review.  Combustion and Flame, 42: 215 —242.  Chen, Z. Y., and Zhao, J. P. (2000). Optical constants of tetrahedral amorphous films in the infrared region and at a wavelength of 633 nm, 3. Applied Physic. 87, 4268— 4273.  Chen, Y., Shah, N., Braun, A., Huggins, F.E., & Huffinan, G.P. (2005). Electron microscopy investigation of carbonaceous particle matter  Choi, M. Y., Muiholland, G. W., Hamins, A., and Kashiwagi, T. (1995). Comparisons of the soot volume fraction using gravimetric and light extinction techniques. Combustion Flame, 102, 161—169.  Choi, S., Lee, K.-R., Oh, S.-B., and Lee, S. (2001). Wide wavelength-range optical studies of hydrogenated amorphous carbon films: from 700 mu to 10 m. Applied Surface Science, 169, 217—222.  Chhowalla, M. , Ferrari, A. C., Robertson, J., & Amaratunga, G. A. (2000) Evolution of sp2 bonding with deposition temperature in tetrahedral amorphous carbon studied by Raman spectroscopy. Applied Physics Letter, 76, 1419  —  1421.  110  Clague, A.D., Donnet, J.B., Wang, T.K., & Peng, J.C. (1999). A comparison of diesel engine soot and carbon black, Carbon, 37, 1553  —  1565.  Cleary, T., Sampson, R., & Gentry, J.W.(1990). Methodology for fractal analysis of combustion aerosols and particle clusters. Aerosol Science and Technology, 12, 518  —  525.  Colbeck, 1., Appleby, L., Hardman, E. J., & Harrison, R. M. (1990). The optical properties and morphology of cloud processed carbonaceous smoke. Aerosol Science, 21, 527—538.  Cook, 3., & Highwood, E. J. (2003). Climate response to tropospheric absorbing aerosol in an intermediate general-circulation Model.  Q. J. R. Meteorological Society in press.  Cuesta, A., Dhamelincourt, P., Laureyns, 3., Martinez-Alonso, A., & Tascon, J.M. (1994). Raman microprobe studies on carbon materials. Carbon, 32, 1523 —1532.  Cullity, B.D., & Stock, S.R. (2001). Elements of X-ray diffraction. Don Mills, ON: Addison-Wesley Publishing Inc.  Dana, P., Sylvie, B., Sonia, T., Guy, A., & Günter, S. (2006). A morphological investigation of soot produced by the detonation of munitions. Chemosphere, 65, 821831.  Deviney, M.L., & O’Grady T.M. eds (1976) Petroleum derived carbons. Washington, D.C.: American Chemical Society.  di Stasio, S. (2001). Observation of restructuring of nanoparticle soot aggregate in a diffusion flame by static light scattering. Aerosol Science, 32, 509  —  524.  111  Dippel, B., & Heintzenberg, J. (1999). Soot characterisation in atmospheric particles from different sources by NIR FT Raman spectroscopy. Aerosol Science, 30, 907  —  908.  Dobbins, R. A., & Megaridis, C. M. (1991). Absorption and scattering of light by polydisperse aggregates. Applied Optics, 30, 4747  —  4754.  Donnet, J.B., & Voet A. (1976). Carbon black: physics, chemistry, and elastomer reinforcement. New York, N.Y.: Dekker.  Donnet, J.B., Bansal, R.C., & Wang, M.J. eds (1993). Carbon black: science and technology. New York, N.Y.: Dekker.  Emmerich, F.G. (1995). Application of a cross-linking model to the Young’s modulus of graphitizable and non-graphitizable carbons. Carbon, 33, 47-50.  Farias, T. L., KoylU, U. 0., Faeth, G.M., & Carvalho, M. G. (1995). Computational evaluation of approximate Rayleigh-Daybe-Gans/fractal-aggregate theory for the absorption and scattering properties of soot. Heat Transfer, 117, 152  —  159.  Farias, T. L., Koylu, U. 0., & Carvaiho, M. G. (1996). Range of validity of the Rayleigh-Debye-Gans theory for optics of fractal aggregates. Applied Optics, 35, 6560— 6567.  Ferrari, A. C., & Robertson, J. (2000). Interpretation of Raman spectra of disordered and amorphous carbon. Physical Review B: Condensed Matter and Materials, 6 1,14095— 14107.  Flagan, R. C. (2001). Electrical techniques, in aerosol measurement: principles, techniques, and applications, edited by P. A. Baron and K. Willeke. New York: John Wiley.  112  Forrest, S.R., & Witten, T.A. (1979). Long-range correlation in smoke-particle aggregates. Journal of Physics A: Mathematical and General, 12, 109— 117.  Frenklach, M., Clary, D., Gardiner, W., & Stein, S. (1985) Detailed kinetic modeling of soot formation in shock-tube pyrolysis of acetylene. Proceedings of the Combustion Institute, 20, 887—901.  Frenklach, M., Clary, D., Yuan, T., Gardiner, W., & Stein, S. (1986). Mechanism of soot formation in acetylene-oxygen mixtures. Combustion Science and Technology, 50, 79  —  115.  Frenidach, M., & Wang, H. (1991). Detailed modeling of soot particle nucleation and growth. Proceedings of the Combustion Institute, 23, 1559— 1566.  Fujimoto, H. (2003). Theoretical X-ray scattering intensity of carbons with turbostratic stacking and AB stacking structures. Carbon, 41, 1585-1592.  Gaydon, A.G, & Fairbairn, R. (1954). Fifth international symposium on combustion. Pittsburgh, Penn.: The Combustion Institute, 324.  Gruber, T., Waldeek-Zerda. T., Gerspacher, M. (1994). Raman studies of heat-treated carbon blacks. Carbon, 32, 1377  —  1382.  Dwaze, P., Schmid, 0., Annegarn, H.J., Andreae, M.O., Huth, J., & Helas, G. (2005). Comparison of three methods of fractal analysis applied to soot aggregates from wood combustion. Aerosol Science, 37, 820— 838.  Etter, D. M., & Kuncicky, D. C. (1996). Introduction to MATLAB for engineers and scientists. Upper Saddle River, N. J.: Prentice Hall.  113  Harling, D.F., & Heckman, F.A. (1969). New information on carbon black fine structure. Materie Plastiche ed Elastomeri, 35, 80— 84.  Heckman, F.A. (1964). Microstructure of carbon black, Rubber Chemistry and Technology, 3, 1245—1298.  Heywood, J. (1988). Internal Combustion Engine Fundamentals. New York, N.Y.: Dekker.  Haywood, J. & Boucher, 0. (2000). Estimates of the direct and indirect radiative forcing due to tropospheric aerosols: a review. Reviews of Geophysics, 38, 513  —  543.  Hansen, J. E., Sato, M., & Ruedy, R. (1997). Radiative forcing and climate response. Geophysical Research, 102, 6831  —  6864.  Herd, C.R., McDonald, G.C., & Hess, W.M. (1991). Morphology of carbon-black aggregates: fractal versus euclidean geometry. Rubber Chemistry and Technology, 65, 107—129.  Homann KH. (1998) Fullerenes and soot formation new pathways to large particles in -  flames. Angewandte Chemical International Edition, 37, 2434—2451.  Homann, k., & Wagner, H. (1967). New aspects of the mechanism of carbon formation in premixed flames. Eleventh international symposium on Combution, 317-319.  Hurt, R. H., Crawford, G. P., & Shim, H-S. (2000). Equilibrium nanostructure of primary soot particles. Proceedings of the Combustion Institute, 28, 2539 —2546.  IPCC (2001). Climate change 2001: Scientific basis. third assessment of the intergovermental panel on climate change. New York, N.Y.: Cambridge University Press.  114  IPCC (2005). Report of the IPCC expert meeting on emission estimation of aerosols relevant to climate change. Retrieved December 8, 2006, from www.ipcc.chlmeetings/session24/inf4.pdf.  Ishiguro, T., Takatori, Y. and Akihama, K. (1997). Microstructure of diesel soot particles probed by electron microscopy: first observation of inner core and outer shell. Combustion and Flame, 108, 231 234. -  Ivieva, N. P., MeKeon, U., Niessner, R, & Pöschl, U. (2007). Raman microspectroscopic analysis of size-resolved atmospheric aerosol particle samples collected with an ELPI: soot, humic-like substances, and inorganic compounds. Aerosol Science and Technology, 41,655—671.  Jawhari, T., Roid, A., & Casado, J. (1995). Raman spectroscopic characterization of some commercially available carbon black materials. Carbon, 33, 1561—1565.  Johnson, B.T. (2003). The semi-direct aerosol effect. Ph.D Thesis, The University of Reading.  Johnson, R. I., Jitendra, J. S., Cary, R. A., & Huntzicker,J. J. (1981). An automated thermal-optical method for analysis of carbonaceous aerosol. ACS Symposium Series No. 167, Atmospheric Aerosol: Source/Air Quality Relationships. Washington, D.C: American Chemical Society.  Jones, H. (2004). Source and characterization of particulate matter from a pilot-ignited natural gas fuelled engine. M.A.Sc Thesis: University of British Columbia.  Jullien, R., & Botet, R. (1987). Aggregation and fractal aggregates. Singapore: World Scientific Publishing Co.  115  Jung, H., Kittelson, D. B., & Zachariah, M. R. (2006). Characteristic of SME biodiesel fueled diesel particle emission and the kinetics of oxidation. Environment Science and Technology, 30,4949 4955. -  Kaufman, Y.J., & Koren, I. (2006). Smoke and pollution aerosol effect on cloud cover. Science, 313, 655—658.  Kamimoto, T., Shimono, M., & Kaze, M. (2007). Dynamic measurements of soot aggregate exhaust by a light scattering method. Physics: Conference Series, 85, 1- 6. Kaye, B. H. (1984) Multifractal description of a rugged fine particle profile. Particle Characterization, 1, 14.  Keller, A., Kovacs, R., & Homann, K.-H. (2000). Large molecules, radicals, ions, and small soot particles in fuel-rich hydrocarbon flames. Physical Chemistry Chemical Physics, 2, 1667—1675.  Koren, I., Kaufman, Y. J., Remer, L. A., & Martins, J.V. (2004) Measurement of the effect of Amazon smoke on inhibition of cloud formation. Science, 303, 1342 —1345. Koren, I. (2006). On the importance of locality when estimating the aerosol effects on climate. Retrieved July 12, 2007, from http:/fscitizen.comlscreens/blogPage/viewBlogfsw_viewBlog.php?idTheme= 1 3&idContr ibution= 165.  Köylü, U.O., & Faeth, G. M. (1992). Structure of overfire soot in buoyant turbulent diffusion flames at long residence times. Combustion and Flame, 89, 140— 156.  KOylu, U. 0., and Faeth, G. M. (1994). Optical properties of overfire soot in buoyant turbulent diffusion flames at long residence times. Heat Transfer, 11, 152—159.  116  KOylu, U. 0., & Neer, A. (2006). Effect of operating conditions on the size, morphology, and concentration of submicrometer particulates emitted from a diesel engine. Combustion and Flame, 146, 142  —  154.  Koylu, U.O., Faeth, G. M., Faria, T.L., & Carvaiho, M.G. (1995). Fractal attd projected structure properties of soot agrregates. Combustion and Flame, 100, 621 —633.  Krestinin, A. (2000). Detailed modeling of soot formation in hydrocarbon pyrolisis. Combustion and Flame, 121, 513  —  524.  Krishnan, S.S., Lin, K.C., & Faeth, G.M.. (2001). Extinction and scattering properties of soot emitted from turbulent diffusion flames. Heat Transfer, 123, 331—339.  Kshirsager, A.M. (1972). Multivariate Statistical Methods. New York, N.Y.: Dekker.  Lahaye J., & Prado G. (1978). Chemistry and physics of carbon: vol. 14. New York N.Y.: Decker.  Lee, K. 0., Cole, R., Sekar, R., Choi, M. Y., Kang, J. S., Bae, C. S., and Shin, H. D. (2002). Morphological investigation of the microstructure, dimensions, and fractal geometry of diesel particulates. Proceedings of the Combustion Institute, 29, 647  —  653.  Lee, K. 0, Cole, R., Sekar, R., Choi, M. Y., Kang, J. S., Bae, C. S., and Zhu, J. (2001). Detailed characterization of morphology and dimensions of diesel particulates via thermophoretic sampling. SAE Technical Paper, (2001-01-3572).  Lehmpuhl, D. W. , Ramirez-Aguilar, K. A., Michel, A. E., Rowlen, K. L., & Birks, J. W. (1999). Physical and chemical characterization of atmospheric aerosols by atomic force microscopy. Analytical Chemistry, 71, 379 —383.  117  Lower. S. (2007). The hybrid orbital model  —  part 1. Retrieved January 15, 2008, from  http://www.chem 1 .comlacad!webtextlchembondlcbO6.html.  Marshall J., Lohman, U, Leaitch, W.R., Shantz, N, Phinney, L., Toom-Sauntry, D., & Sharma, S. (2005). Optical properties of aerosol particles over the northeast pacific. Applied Meteorology, 44, 1206  —  1220.  Martin, G. M., Johnson, D. W., & Spice, A. (1994). The measurement and parameterisation of effective radius of droplets in warm stratocumulus clouds. Atmospheric Science, 51, 1823  —  1842.  Mathis, U., Mohr, M., & Kaegi, R. (2005). Influence of diesel engine combustion parameters on primary soot particle diameter. Environmental Science and Technology 39, 1887—1892.  McTaggart-Cowan, G.P., Bushe W.K., Munshi, S.R. & Hill, P.G. (2003). A supercharged heavy-duty single cylinder research engine for high-pressure direct injection on natural gas. International Journal of Engine Research, 4, 315  —  330.  Meakin, P. (1983). Formation of fractal clusters and networks by iffeversible diffusionlimited aggregation. Physical Review Letters, 51, 1119  —  1122.  Meakin, P. (1998). Fractal, scaling and growth far for equilibrium. Cambridge: Cambridge University Press.  Megaridis, C. M., & Dobbins, R. A. (1990). Morphological description of flamegenerated materials. Combustion Science and Technology, 71, 95— 109.  Mernag,  t. Ii., Cooney, 1k.. P., & Johnson, R. A. (1984). aaman spectra of graphon carbon  black. Carbdh 1984, 22, 39—42.  118  Mikromash. (2008). HOPG: Highly Ordered Pyrolytic Graphite. Retrieved December 1, 2007 from http://www.spmtips.com/hopg/.  Mitchell, P., & Frenklach, M. (1998). Monte carlo simulation of soot particle aggregation with simultaneous surface growth  —  why primary particles appear spherical. Proceedings  of the Combustion Institute, 27, 1507— 1514.  Moskieva, L., Mebel, A., & Lin, M. (1996) The CH 3  +  H reactions: a potential source 5 C  of benzene at high temperatures. Twenty-six Symposium (International) on Combustion, 521 —526.  Mikhailov, E.F., Viasenko, S.S., Kiselev, A.A., & Ryshkevich, T.I. (1998). Restructuring factors of soot particles. Atmospheric and Ocean Physics, 34, 307-3 17. Muiholland, G.W., Bobre, C.F., & Fuller, K.A. (1994). Light scattering by agglomerates —  coupled electric and magnetic dipole method. Langmuir, 10,2533  —  2546.  NASA (1999). The NASA facts: aerosols. Retrieved October 11, 2007, from gsfc.nasa.gov/gsfc/. ./gallery/fact_sheets/earthscilterralaerosols.pdf. .  Naydenova, 1.1. (2007). Soot formation modeling during hydrocarbon pyrolysis and oxidation behind shock waves. Ph.D Thesis, Rupertus Carola University of Heidelberg. NIOSH (1999). NIOSH manual of analytical method  —  fourth edition. Retrieved June 11,  2007, from www.cdc.gov/nioshlnmamlpdfs/5040.pdf.  Oberlin, A. (1984). Carbonization and graphitization. Carbon, 22, 521 —541. Oktem. B.. Tolocka, M. P.. Zhao. B., WanQ, H.. Johnston. M.V. (2005) Chemical species associated with the early stage of soot growth in laminar premixed ethyleneoxygen argon flame. Combustion and Flame 142, 364  —  373.  119  Onisehuk, A.A., di Stasio, S., Karasev, V.V., Baklanov, A.M., Makhov, G.A., Vlasenko, A.L., Sadykova, A.R, Shipovalov, A.V., & Panfllov ,V.N. (2003). Evolution of structure and charge of soot aggregates during and after formation in a propane/air diffusion flame. Aerosol Science, 34, 383—403.  Palas. (2003). Operating manual: graphite aerosol generator GFG-l000.  Palmer, H., & Cullis, H. (1965). The chemistry and physics of carbon: vol.1. New York, N.Y.: Dekker.  Park, S. H. and Rogak, S. N. (2003) A one-dimensional model for coagulation, sintering and surface growth of aerosol agglomerates. Aerosol Science and Technology, 37, 947  —  960.  Popovitcheva, 0. B., Persiantseva, N. M., Trukhin, M. E, Rulev, G. B., Shonija, N. K., Bunko, Y., Starik, A.M., Demirdjian, B.  ,  Ferry, D.,& Suzanne, J. (2002). Experimental  characterization of aircraft combustor soot: microstructure, surface area, porosity and water adsorption. Chemical Physics and Physical Chemistry, 2, 4421- 4426.  Porter, G. (1955). Fourth conbustion symposium. Cambridge, Mass, 248.  Prakash, C.A. (1998). A critical review of biodiesel as a transportation fuel in Canada. Environment Canada.  Ramanathan, V., Crutzen, P. J., Kiehi, J. T., Rosenfeld, D. (2001). Aerosols, climate, and the hydrological cycle. Science, 294, 2119  —  2124.  Ravier, J. Houze, F., Carmona, F., Schneegans, 0., & Saadaoui, H. (2001). ,  Mesostructure of polymer/carbon black composites observed by conductive probe atomic force microscopy. Carbon, 39, 314—318.  120  Robertson, J. (1986). Amorphous carbon. Advance Physics, 35, 3 17—374.  Robertson, J. (1992). Mechanical properties and coordinations of amorphous carbons. Physical Review Letters, 68,220—223.  Roessler, D. M., and Faxvog, F. R. (1981). Changes in diesel particulates with engine air/fuel ratio, Combustion Science and Technology, 26, 225 —231.  Rogak, S.N., & Flagan, R.C (1992) Characterization of the structure of agglomerate particles. Particle & Particle Systems Characterization, 9, Mar, 19 27. -  Rogak, S.N., Flagan, R.C., & Nguyen H.V. (1993). The mobility and structure of aerosol agglomerates. Aerosol Science and Technology, 18, 25  —  47.  Romanou, A., Liepert, B., Schmidt, G. A., Rossow, W. B., Ruedy, R. A., & Zhang, Y. (2007). 20th century changes in surface solar irradiance in simulations and observations. Geophysical Research Letters, VOL. 34, L05713.  Rosen, H., & Novakov, T. (1977) Raman scattering and the characterization of atmospheric aerosol particles. Nature, 266, 708  —  710.  Rosen, Fl., & Novakov, T. (1978). Identification of primary particulate carbon and sulfate species by Raman spectroscopy. Atmospheric Environment, 12, 923 —927.  Sadezky, A., Muckenhuber, H., Grothe, H., Niessner, R., & Pöschl, U. (2005). Raman spectra of soot and related carbonaceous materials: spectral analysis and structural information. Carbon, 43, 1731  —  1742.  121  Saito. K., Gordon, A. S., Williams, F. A., and Stickle, W. F. (1991). A study of the early history of soot formation in various hydrocarbon diffusion flames. Combustion Science and Technology, 80, 103  —  119.  Sakurai, H., Park, K., MeMurry, P. H., Zarling, D. D., Kittelson, D. B., & Ziemann, P. J. Size-dependent mixing characteristics of volatile and nonvolatile componenets in diesel exhaust aerosols. Environmental Science and Technology, 37, 5487  —  5495.  Samson, R.J., Muiholland, G.W., & Gentry, J.W. (1987). Structural analysis of soot agglomerates. Langmuir, 3, 272—281.  Satheesh, S. K., & Ramanthan, V. (2000). Large differences in tropical aerosol forcing at the top of the atmosphere and earth’s surface. Nature, 405, 60—63.  Sato, S., Kawamata, K., Kurumada, A., Kawamata, M., & Ishida, R. (1988). A rationalization of the graphitizing heat treatment process of carbon. Carbon, 26, 465-476 Scanlon, J.C., & Ebert, L.B. (1993). X-ray diffraction study of fullerene soot Source: Journal of Physical Chemistry, 97, 7138-7140.  Schnaiter, M., Horvath, H., Mohler, 0., Naumann, K.-H., Saathoff, H., and Schock, 0. (2003). UV-VIS-NIR spectral optical properties of soot and soot-containing aerosols, Aerosol Science, 34, 1421  —  1444.  Schuetz, C. A., & Frenklach, M. (2002). Nucleation of soot: molecular dynamics dimulations of pyrene dimerization. Proceedings of the Combustion Institute, 29, 2307  —  2314  Schraml S. Dankers S, Bader K, Will 5, & Leipertz A. (2000). Soot temperature measurements and implications for time-resolved laser-induced incandescence. Combustion Flame, 120 439 ,  —  450.  J 22  Seinfeld, J. (1986). Atmospheric chemistry and physics of air pollution. New York, N.Y.: John Wiley & Sons, Inc..  Seinfeld, J., & Pandis, S. (1998). Atmospheric chemistry and physics: from air pollution to climate change. New York, N.Y.: John Wiley & Sons, Inc..  Shaddix, C. R., Palotas, A. B., Megaridis, C.M., Choi, M.Y., & Yang, N.Y. 2005. Soot graphitic order in laminar diffusion flames and a large-scale JP-8 pooi fire. Heat and Mass Transfer, 48, 3604  —  3614.  Shim, H-S, Hurt, R.H., & Yang, N. (2000). Methodology for analysis of 002 lattice fringe images and its application to combustion-derived carbons. Carbon, 38, 29  —  45.  Skillas, G., Kiinzel, S., Burtscher, H., Baltensperger, U., & Siegmann, K. (1998). Hihh fractal-like dimension of diesel soot agglomerate. Aerosol Science, 29, 411  Smallwood, G. J., Snelling, D. R., Liu, F., & Gtilder,  -  419.  O. L. (2201). Clouds over soot  evaporation: errors in modelling laser-induced incandescence of soot. Heat Transfer, 123, 8 14-818.  Song, J., Alam, M., Boehman, A.L.,  *  Kim, U. (2006). Examination of the oxidation  behaviour of biodiesel soot. Combustion and Flame, 146, 589  —  604.  Sorensen, C. M. (2001). Light Scattering by Fractal Aggregates: A Review, Aerosol Scence and. Technology, 35, 648  —  687.  Sorensen, C. M., and Feke, G. D. (1996). The Morphology of macroscopic soot, Aerosol Combustion Science and Technology, 25, 328  —  337.  Strapp, J. W., Leaitch, W. R., & Liu, P. 5. (1992). Hydrated and dried aerosol size dispjJ,ution measurements from the Particle Measuring Systems FSSP-300 probe and the  123  deiced PCASP-100X probe. Atmospheric and Oceanic Technology, 9, 548—555. Svensson, K. (2005). Effects of fuel molecular structure and composition on soot formation in direct-injection spray flames. Ph.D Thesis, Brigham Young University. Sze, S. K., Siddique, N., Sloan, J. J., & Escribano, R. (2001). Raman spectroscopic characterisation of carbonaceous aerosol. Atmospheric Environment, 35, 561—568. Tanner, S.D., Goodings, S.M., & Bobme, D.K. (1981) Hydrocarbon ions in fuel-rich, methane-acetylene-oxygen flames as a probe for the initiation of soot: an experimental approach. Canadian Journal of Chemistry, 59, 1810-1818.  Tence, M., Chevalier, J. P., & Jullien, R. (1986). On the measurement of the fractal dimension of aggregated particles by electron microscopy: experimental method, corrections and comparison with numerical models. Journal de Physique (Paris), 47, 1989-1998.  TSI (2003). Model 3936 SMPS instruction manual. Revision H: September 2003. Tuinstra, F., & Koenig, J.L. (1970). Raman spectrum of graphite. Chemical Physics, 53, 1126—1130.  Twomey, S. A. (1977). The influence of pollution on the shortwave albedo of clouds. Atmospheric Science, 34, 1148  —  1152.  Vander Wal, R.L., & Tomasek, A.J. (2003). Soot nanostructure: dependence upon synthesis condition, Combustion and Flame, 136, 129  Vander Wal,  —  141.  k. L., & TolrLasek, A. 3. (2004). Soot nanostructure: defmition,  quantification and implication. SAE Technical Paper, (2005-01-0964).  124  VanderWal, R. L., Tomasek, A. J. Pamphlet, M. I., Taylor, S. D., & Thompson, W. K. (2004). Analysis of HRTEM images of carbon nanostructure quantification. Nanoparticle Research, 6, 555  —  568.  Violia, A., & Venkatnathan, A. (2006). Combution-generated nanoparticles produced in benzene flame: a multiscale approach. Chemical Physics, 125, 1-5  Wang, Y., Alsmeyer, D.C., & McCreery, R. L. (1990). Raman spectroscopy of carbon materials: structural basis of observed spectra. Chemistry and Material, 2, 557—563.  Weingartner, E., Burthscher H., & Baltensperger, U. (1997) Hygroscopic properties of carbon and diesel soot particles. Athmospheric Environment, 31, 2311  —  2327.  Williams, A., McCormick, R. L., Hayes, R. R., Ireland, 3., & Fang, H.L. Effect of Biodiesel Blends on diesel Particulate Filter Performance. Retrieved September 10, 2006, from www.nrel.gov/docs/fyo7osti/400 15 .pdf.  Zhang, H. (2005).Nitrogen evolution and soot formation during secondary coal pyrolisis. Ph.D Thesis, Brigham Young University  Zhu, J., Lee, K. 0., Yozgatligil, A., & Choi, M.Y. (2005) Effects of engine operating conditions on morphology, microstructure, and fractal geometry of light-duty diesel engine particulates. Proceedings of the Combustion Institute, 30, 278 1—2789.  125  APPENDIX A THERMOPHORETIC SAMPLER  Transmission Electron Microscopy (TEM) has been frequently used to determine the morphology, composition, and nanostructure of aerosol particles deposited on metallic grids. However, with resolution of the TEM on the order of tenths of nanometers, aerosol particles must be intelligently deposited on the grid to ensure adequate coverage without overlapping. For aerosol research, samples are usually collected on a thin copper grid coated with silicon oxide or a carbonaceous material  Particles may be deposited onto a TEM grid after being collected on a filter. This method is used for inorganic fibers such as asbestos (Sahle & Ildiko, 1996). The filter is either dissolved or partially oxidized with low temperature plasma to allow the particles to be deposited on the TEM grid. However, this method will not work with aerosols that are susceptible to oxidation or dissolution.  Impaction has also been used to collect aerosol particles on TEM grids. Pósfai et al  (1995) used a TEM grid placed on the third stage of a cascade impactor. Jones (2004) collected soot samples for nanometer characterization on a lacey formvar coating by passing a very low flow of exhaust gas through thegrid for 5-10 minutes. Although impaction was effective for particles over 100 nni, the collection efficiency decreased substantially as the smaller particles tended to follow the sttedhilines (Jones, 2004).  126  Electrostatic samplers allow charged aerosol particles to beefficiently deposited on a flat collection plate by using an electric field. Recently, Fierz et al. (2007) presented a portable electrostatic TEM sampler for use in ambient air studies. Again, however, the smaller nanoparticles (<30 nm) had small charging efficiencies and thus poor collection efficiencies for electrostatic precipitators (Dixkens & Fissan, 1999).  In thermophoresis, there is a tempeiature gradient in the entraining gas between two surfaces. More high energy gas molecules collide with the particles on the hot surface side than on the cool surface, resulting in a net force on the particle towards the cooler surface (Seinfeld & Pandis, 1998). Thermophoresis has been used for aerosol sampling since Green and Watson employed a heated nichrome wire to measure dust concentrations in British mines (Gonzales et al., 2005). Heated wire thermophoresis has been used for more current studies; however, the high local temperature around the wire may cause chemical reactions to occur to the aerosol. Dobbins and Megaridis (1991) pioneered the use of thermophoresis for TEM support grids by quickly inserting a cooled probe and TEM grid parallel to the flow of gas using a pneumatically actuated insertion device. This method has been used for both engine exhaust (Neer & KOylu 2006) and fundamental flame studies (Hu & Koylti, 2004). Plane-to-plane precipitators have also been used for attaining more uniform particle deposition (Gonzales, 2005).  Point-to-plain thermophoretic precipitators have also been discussed, either through a cooled sampling plate (Dixkens & Fissan, 1999), or through a heated jet (Lee & Kim, 2003). With point to plane deposition, much larger temperature gradients can be  127  achieved over the smaller area of standard size TEM grid. This paper presents a new heated jet point-to-plane thermophoretic sampler that heats the sample jet with electrically resistive hypodermic tubing.  The thermophoretic sampler (TPS) used in 1his study was designed to be robust in order to provide accurate reproducible deposition. Figure A.1 shows a design of the TPS. The dimensions are 10cm x 5 cm diameter connected to ¼” Swagelok fittings at both the  ground electrode and the positive electrode.  Resistance heating is used to heat up a small hypodermic tube that heats the sample gas as it passes through it. Hypodermic tubing that is sufficiently thin is needed in order to increase the resistance of the tubing so that heat can be produced at low current. For the current design, a 0.02” ID x 0.025” OD x 2.5” hypodermic tube is used. The tubing is press fit into a small hole drilled in 1/16” sheet metal which acts as the positive electrode. Due to the relatively small area of heating, the contact area between the hypodermic tubing and the positive electrode is kept to a minimum in order to ensure a low temperature drop at the end of the tubing. At least 3 cm length is needed in order to ensure that the sample gas is sufficiently heated. The heating tube was insulated with Teflon in order to ensure lower heating requirements. The temperature gradient is varied by varying the current that passes through the hypodermic tubing.  128  I  SN,PLE IT GAT1VE  ftECT1OOE/ 0.025Co K o25- I) STAI’LESS 0T  T HOER  Figure A.1 Teflon TPS with Brass TEM Holder  The TEM grid is held in place with an annular brass cap on a brass stage TEM holder. The brass stage TEM holder ensures that the TEM grid remains cool during the sampling time. The brass stage that holds the TEM grid is assumed to be near ambient temperature. A distance of 5mm between the jet outlet and the grid holder ensures that all of the particles that come in contact with the 3mm grid have come directly from the heated jet. Stage distances that are greater than 10 times the nozzle diameter has been shown to lose significant amounts of momentum before impinging on the’ plate surface (Lee & Lee, 1999).  The TEM holder shown in Figure A. 1 can be loosened and removed from the TPS. A small recess in the stage ensures the TEM grid is centred on the grid. After replacing the cap on the stage, the TEM holder assembly is tightened back in the TPS. The setup as  129  shown in Figure 1 shows that the flow over the TEM should be axisymmetric about the centre. Unfortunately, with the current design it is difficult to accurately centre the heated jet on the grid. The 0.025” hole through the electrode must be perpendicular to the electrode face. The two screws holding the positive electrode must be concentric with the 0.025” hole drilled through the electrode. Any misalignment between the Teflon  insulation, the positive electrode, and the TEM grid holder will result in maximum deposition from the heated jetto be off of the TEM centre.  A.2  References  Dixkens, J., & Fissan, H. (1999) Development of an electrostatic precipitator for off-line particle analysis. Aerosol Science and Technology, 30, 438-453.  Dobbins,R. A., & Megaridis, C. M. (1991). Absorption and scattering of light by polydisperse aggregates. Applied Optics, 30, 4747  —  4754.  Fierz, M., Kaegi, R., & Burtscher, H. (2007) Theoretical and experimental evaluation of a portable electrostatic TEM sampler. Aerosol Science and Technology, 41, 520—528.  Gonzalez, D., Nasibulin, A.G., Anatoli, M. Baklanov, Shandakov, S.D., Brown, D. P., Queipo, P., & Kauppinen, E. (2005). A new thermophoretic precipitator for collection of nanometer-sized aerosol particles. Aerosol Science and Technology, 39, 1-8.  Hu, B., & Koylu, U.O. (2004). Size and morphology of soot particulaes sampled from a turbulent non-premixed acetylene flame. Aerosol Science and Technology, 38, 1009— 1018. Jones, H. (2004). Source and characterization of particulate matter from a pilot-ignited natural gas fttl1ed ëfigine. MASc Thesis, University of British Columbia.  I ii  Lee, J., & Lee, S-J. (1999) Stagnation region heat transfer of a turbulent axisymmetric jet impingement. Experimental Heat Transfer, 12, 137-156.  Lee, U.L., & Kim, S.S. (2002). The effect of varying impaction plate temperature on impactor performance: experimental studies. Aerosol Science, 33, 451—457. Neer, A., & Koylti, U. O.(2006). Effect of operating conditions on the’ size, morphology, and concentration of submicrometer particulates emitted from a diesel engine. Combustion and Flame, 146, 142  —  154.  POsfai, M. , Anderson, J., & Buseck, P. (1995)ompositional variations of sea-salt-mode aerosol particles from the North Atlantic. Geophysical Research, 100, 23063 —23074.  Sahle, W., & Ildiko, L. (1996). Airborne inorganic fibre level monitoring by transmission electron microscope (TEM): comparison of direct and indirect sample transfer. Annals of Occupational Hygiene, 40, 29-44.  Seinfeld, J.H., & Pandis, S.N. (1998). Atmospheric chemistry and science: from pollution to climate change. New York, N.Y.: John Wiley & Sons, Inc.  131  APPENDIX B SAMPLE DATABASE FILES  Original Image  Box C Grid 75 Print May: 237000x 75mm 1326 05/23/07 Microecopist: Arka  20 ma HV=8OkV Direct Hag 500000x X: .9 5’: 25.3 ubc bioimagin  Figure B.1 The TEM Original Image  132  Cropped and Filtered Image  Figure B.2 The Cropped and Filtered Image  Binary Image  Figure B.3 The Binary Image  133  Edge Image  Figure B.4 The Edge Image  Final Imposed Image  Figure B.5 The Final Imposed Image  134  Length and Width Image  Figure B.6 The Length (Blue) and Width (Red) Image  Primary Particle Image  Jr  —-  r  ,‘.  I  Figure B.7 The Length (Blue) and Width (Red) of the Primary Particles  135  Density-Density Correlation Plot Density-Density Correlation Function I  I  I  I  I  :.:  Oo  1b  Fiaure B.8 The Density-Density Correlation Fun:  -  Pair Correlation Function Plot Pair Correlation Function I  I  0  EN 0.5  0.3  0.1 0  0  I  I  I  I  20  40  50  ftl  100  120  140  Figure B.9 The Pair Correlation Function 136  Density-Density Correlation and Pair Correlation Plot The Pair Correlation and Density-Density Correlation Functions I  1  I  I  I  I  PCM DDC  0.9 0.8 0.7 0.8  0.3 0.2  0  I  I  I  20  40  60  80  100  120  140  Figure B.1O The Comparison between Density Correlation and Pair Correlation  ii,  Database Text File =============ADMINI STPJ\TION============= TEM Session Date (dd/rnm/yy) : DEC. Bio±maging Grid ID: 7A Engine: VW TDI Enaine Mode: I Engine Test Date (dd/mm/yy) : Fuel: ULSD Magnification: 400kx  23/05/07  Location:  Ti  VOiCdye;  18/12/06  OUJ’V  Number of Agglomerates in original Image: 2 Image Name: D:\Graduate UBC\HRTEM\Box C\Rogak Box C Grid 7A\Rogak Box C ,—,  __77  (_r\1  -ç  =====—===——=======ANALYS IS The AQqlomerate of Number: Scale Factor [nm/pixel] Length [nm] = 137.679 = vidth [nat] Diameter [flirt] = 38.6 Perimeter [nm] = 635 Area Fnm*21 5195.21  =  1 0.3125  85  Density-Density Correlation: There are 319 data points. r C(r) 0 0.999999 1 0.980282 1 0.970178 1 0.964116 2 0.955023 0.945741 2 u.938092 3 0.93133 3 0.922144 3 0.909733 3 0.89845 4 0.889291 4 0.881895 4 0.874166 0.867576 5 0.860748 5 5 0.85452 0.846612 6 6 0.83855 6 0.830578 7 0.822838 7 0.814395 7 0.804947 8 0.796466  138  )— F—i F—i I— I—j F—a f—i FF- cc) c cø C) 0) CX) 03 0) 0) 0) N) N) N) I- I- i- 00)0  0) CT’ CT’ 0 0) cY 0 —-.1 0 —3 —3 0) CCI I—i 0.) —.3 N) C) C) C) Cii  i—i  01 0) C) C)  (ii —3 C) N) C)  C) C— C) Cii c  C) C) CT’ C) cc)  —-.3 — C) cc) 0) —3 C) C) c) co 01 F— —.1 0) cc> C) 0 X C) 0 C) 0) C) 0) CX) c> cc> C) C)  CT’ 0 01 —.3 0 0)  C)] CC] I—i N) C) C) C) 0] 0) —.3 C) CT’ N) 0) cc> 0) cc>  01 N) cc) cc> —1  Cii 0) 01 c 0] C) Cii 01 . NJ C) F—3 i—i —3 C)  01 01 0) —1 0) CC]  01 0] cc> —3 N) U) (ii C) Cii I—a c) N) CT’ —3 — 01 I—i 0]  C)) -i —3 C) 01 N) 01 C) 0) cc) N) C)  Cl] cc> 0 —3 N) c)-  01 0) C) —.3 cc> 0) C) C) 00 — -.1 cc> cc> ui C) H’ N) N) 0) 0 0) C) C) C) C) cc> cc) —1 .X  U) c) 0) 0 —3 H Co C)  c> C) —.3 cc> cc> 01 C) f—i oi C) C) 0) 0) FI-’  —3 —3 0 C) N) CT’ F—’ 01 0)0.)  Co 0) —3 F—’ N) cc> F—.3 . NJ cc> —-3 —.3 N)  0 C) 01 N) N) 0)01 0) C) N) N) C) cc) 0)  N) cc> cc> F-’ F-’  0) C) C) U) F-’  C)  C) 030) C)  i 0-C C) —.3 C) -.3 N) c) cc> -.3 N) C) 0 C) CT’ N) C) NJ N) I’-) kr’— cc>  C) C) C) 0) C) C) C) C) C) C) C) 0_C C) —.3 —.3 —-.1 —3 —-.3 —.3 -.1 —1 —-.1 —.3 —.3  O 0) 0 0 0 0 C) 0) 00 0 0 0 C) 00 C) 0 0 0 0 0 0 C) 0 0 0 0 0 0 0 0 0 0 00000000000000000000000  N) N) N) N) N) N) N) N) N) N) N) N) N) N) N) N) N) N) N) F- F- F—i F- F—i I—i F- F--i [—i F- I—a I t— H F- f—i 01 Cii 01 0) 0) 0) 0) N) N) Ni I—s I- - 000cc) c cc) CX) C) 0) Cc) —.3 —.3 —.3 C) C) C) 01 CT’ CT’  . . ci ci ci ci c c ci ci ci ci ci ci (i ci ci ci ci ci ci c. c. ci ci ci ci ci ci ci ci ci ci ci t\ r’i r’j r-o i r i’i r’ r’i t’j r’j N) F- i- 000 D co GD GD GD GD i —J 1 GD GD GD (ii U’ cri . ci ci W Ui Ni N) Ni I—’ 000 ‘D D GD GD CO GD —J -J —3 GD GD GD  N) —3 U) U) GD GD  P-i —.1 GD N) GD GD  N) N) I’) N) GD GD GD GD 0 N) O GD H GD H F- . 0 —1 (3] U’  P0 CD GD -J t t  N) N) NJ N) U) U) U) U) 0 U) CC) —3 GD N) GD Ui GD U) Ui—i GD F- U) U) —  ci Ui Ui Ui Ui Ui U) U) 0 0 0 0 0 C—a f— 0 N) U’ --i U) Ni . —3 Ui GD NJ H U’ Ui U) GD —3 -i--i U) P-i NJ GD Ui NJ U) U) N) NJ -  —  N) N) N) N) f—i Ui ( Ui U’ U’ (ii U’ GD GD GD —a --i —i —3 GD o —3 C) N) cii U) F- U) GD U) N) . —3 Ui GD U) tO U’ N) U) f-) GD U) CD I—i -J GD GD H 0 CD GD 0 GD GD Ui —3 1—’ GD U’ U) 0 N) U) GD GD —3 Ui N) 0 —3 N) Ui N) GD GD Ui U) U) U) U) GD U) U’ U’ Ui GD GD GD GD GD i GD GD GD —3 Ui U) (3] F-  0 NJ (ii —3 U) 0  o  . —  —  0) P0 U’ U) P0 GD U) Ui —3  0 0 0 to to to C ) 4 GD N) Ui 0 GD U’ C) U) CC —3 Ui f-.) -3 0 U) GD -i--i 0 F- U’ Ui U’ Ui U) GD —3 U’ U) GD GD U’ U’ GD -J U) .t GD  GD U) U)  Ui U) Ui Ui U) Ui U) U) Ui Ui U) Ui Ui Ui Ui U) Ui U) U) Ui U) Ui Ui Ui Ui Ui Ui U) Ui  0 0 0 0 0 0 0 0 0 0 0 0 0 CD CD 0 0 0 0 0 CD 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 C) 0  ci ci ci N) N) N)  F—’ 01 CO N) -3  F—’ 01 ‘.0 F-’ CO CO  CO CO CO CO ON —1 -.3 -J —1 —I Ci) -3(0 0 Cii Ci) CO —3 0 CO (0 . CO I—’ OW F-’ CO ‘.0 0 N) F—’ N) (0010—J0 01 N) 0’ F-’ 01 ‘ CO —1 CO CO -  F—’ I—’ F—’ F—’ F—’ I—’ F—’ F—’ F—’ F—’ F—’ F—’ F—’ F-’ I—’ I—’ F—’ CO 0) CO CO CO co coN) F-’ Cli ‘.00 $ F-’ CO (51(0(0 CO 01 -3(71 NJ C) 10) N) CO CO Cli I—’ N) F-’ I—’ F—’ F—’ F—’ I—’ (‘3 N) NJ N) NJ (0 ‘.0 co ‘.0 ‘.0 00000 —3 CO NJ 3 ‘ CO (0 F-’C) ‘ CO —5 —5 CO (3’ CO —3 0) F—’ CO 0) CO CO CO (0 CO ‘.0 (51 CO CO (0010101 I—’ i Ci) -3 N) F-’ 0) I—’ NJ N)  (‘-3 N) N) I-’ F-’ F-’ . CO CO N) N) CO —3 Cli I-’ NJ CO NJ N) NJ F-’ N) N) (ON) 0) NJ —5 ‘.0 N) -301 CO 01 CO (“3 F’—) NJ N) N) F’-) NJ N) 0) (‘3 0) Ci) 01 (0 CO ‘.00 ON CO F-’ F—3(0 -1(0 (0030 I-’ Ci) ‘.0’.0 01  ‘  0)  3 N) ‘.000) 0 C) —i —5 ‘ N) . ‘ F-’  CO -) -J  N)  o 0 Po  ‘  N) N) N) N) (‘—3 N) NJ N) N) NJ N) NJ ui u-i Cli 01(71 CO CO ON I-’ N) CO —3 0 N) Ci) CO CO N) Ci) CO N) CO —F N) CO N) ‘.0 (OCO 0 ‘.0 —3 I-’ F—’ 0) ‘51 CO -3 C) 01010100 CO (Ti CO 01010’ N) CO N) (51 N)  0 0 00 C) 0 C) 0 0 0 0 C) 0 0 0 0 0 0 0 C) C) 0 0 C) 0 00 C) 0 0 00 Cl 0 0 0 C) 0 C) C) 0 00 0 0 C) 0 0 0 0  CO ON ON ON CO (51 (51 (51 ui u-i u-i u-i (31 U-i (51 (51 Cli Cli Ui (51 (J1 u-i u’ u-i u-i u, (51 (71 (51 u’ u-i Q1 u-i (31 (51 U-i u-i o. . .o. F-’ I-’ 000 (0 (0 (0 CO 0) CO CO —3 —3 -.3 ON CO CO u-i (Ti Cli 0) 0) 0) 0) NJ N) N) F-’ F—’ I—’ C) 00 (0 (0 (0 CO CO (Xi CO —3 —3 —3 CO CO CO Cli Cli (51  NJGD  -  CD 0 C) 0 0 0 0 0 0 0 C) 0 000 —3 -J —3 —3 CC CC 0) CC CC CC CC GD 0) 0) U’ Q] U’ GD CC 0 H’ NJ U’ 0 —3 0)0) NJ U’ H’ NJ NJ U’ 0< NJ 0) I—’ U’ U’ N-) —3 00) H’ NJ H’ 0) U’ U’ 0 -J CC NJ U’ CC 0) 0) 0) 0 H’ —3 F-’ N) LX) N-) U’ 00 U’ H’ GD NJ —3 CC NJ U’ 00 U’ —3 U’ —3 U’ U’ U’ —3 a) U’  C) 0 CD 00 0)0)0) (3) -J —i 0) —3 0) 0) 0 NJ U’ NJ U’ CC C) 0 F 00)0 U’ U’ NJ—ia)•-J  o -  000000000000000000000  0 0 0 H’ H’ F-’ I—’ F-’ H’ F-’ F—’ F-’ I—’ H’ H’ H’ H’ H’ H’ F-’ H’ F-’ H’ F-’ H’ H’ I-’ H’ H’ I-’ I—’ H’ H’ 0) 0) 0) 000000 H’ I—’ I—’ I—’ H’ I—’ NJ NJ NJ NJ NJ NJ U’ U’ U’ U’ U’ ‘ U’ U’ <3) —3 GD I-’ H’ U’ U’ —3 00 U’ .G 0’ CCCCO NJ 0) (3) 00)0 -J 0) H’ U’ U’ CCC) NJ NJ U’ U’ H’ U’ 0<0 CC U’ CC U’ U’ 0 U’ 0) H’ 0 C) —3 0) ‘ 0)00 H’ NJ -J U’ NJ U’ H’ NJ 0) 0) U’ U’ 0) H’ U’ U’ 0) -J GD U’ -J —3 0) 0) —3 00 NJ —1 0) CX) U’ 000 NJ - 0cC’ U’ U’ I-’ CC NJ H’ GD NJ LX) H’GD0) —1 F-’ 00) NJ CC NJ H’ . 3)00 NJ NJ -_j NJ  000000000000000000000000000000  ‘  —3—i  LX)  NJ —3 U’  NJ U’ U’  H’ H’ H’ U’ U’ U’  —J —J —J —J —J —J —J —J —3 —J —3 —J —3 —1 —3 —i —3 —3 —3 —3 —3 —3 —3 —1 —3 —3 —J —3 —1 —1 (3) (3) 0 0 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0 CC CC GD a) —3 —1 —3 0) 0) 0) U’ 0’ U’ . U’ U’ U’ U’ N) NJ NJ I—’ F-’ I—’ 0 C) C) GD GD GD a) CC CC —1 —3 — 0) 0) 0F U’ U’ 01 3 . U’ (.) U’ U’ NJ N) NJ F-’  1-’  000000 NJ NJ 5’) NJ NJ NJ 0). . . (5101 0) C.j —J CO —J 0) 0) NJ . 00) —.1 01 I00(0 0-J 0)00)  00000000000 NJ NJ NJ NJ NJ NJ 5’) 5’) (5) (5) (5) 0)0 —J CO —1 0(0000 NJ 0(0 5’) 5’) 0000000 01 I—’ 0(0 I-’ 0) 0001 F-’ (0(ONJ00’J5’) 05’) 0NJ010) C—’  00000000000 000000(000 0)0) (5) 0) 0) 0) (5) 0) C.) 5’) (5) 0) 0100 -) I-’ 0 NJ 5’) (0 (5) 0101 01 H’ 0) 0’ —J 5’) 00) NJ H’ 01 5’) 00)01 (5) C-’ 0010(51 0000000000 0)0)0) 0) ‘ . .0. .0. .0. .0. 00(0(5) H’ H’ NJ 0)0) U’ 0) 000010 C.) 00) I—’ (51 H’ .0. I-’ —) 0(00) () 0’ 0000)0)010)000) (5) H’c0—J5’) W 000000000000 .0. .0. .0. .0. .0. 01000010 010) —3 ) (0(00 F-’ NJ (5) .0. 01 .0. .0. 010000) NJ H’ CX) H’ 0’ H’ 000000000 F-’ (0 (00)01 H’ 5’) 0)010 NJ 0) 01(0 NJ  NJOWNJH’0  0)  000000000 01000100000 0) —J (0(00 H’ C.) .0. 01 .0. (0 NJ .0. (0 0. (51(510 00)0010 NJ 5’) (3)01 00)0) H’ F-’ (.3 (0 NJ )  000000000000 00000  (0 (0 (0 (0 (0 (0 0 (0 (0 (0 (0 (0 (5) (0 (0(0 (0 (0 (0 0) 0) CO 0) 00) CX) 000000000000000 0000000000 —J -J 0 0 (0 —1 0) 0) 0) 01 01 (51 0) 0) 0) 0) NJ NJ 5’) i- I- I- 0 0 0 (0 (0 (0 CX) (0 CX) CO —J —J —J 0) CX) 0) (51 01 (51 (5) 0) (5) 0)NJ NJ NJ t.H 00 0 (0(0  000 (7101 In H a) (Ti H U) (Ti U) (.51 . a) a) NJ 01 1)1 -J NJ a) F—’ -J 0 H ‘ NJ a) -J U) H I—’ (Ti a) 1.’) U)  a) (0 a) NJ  (V  C-)  a)  H 0 U) 0) 0 NJ 0 0) (0 a) (0 a) 1)1 F-’ NJ .0.  0 U) (0 (0 a) U) a) 0) 0 a) 0 Ui (0 a) (0 . NJ (0  0 NJ U) (ii —1 (0 H  1)101  0) a) a) —.3 —3 a) —J H 0) (0 a) (0 01 0 NJ i- U) NJ 17101 (0 (0 NJ NJ 01—i--i . . U) 1)1 NJ -.1 a) 001 U) a) a) w (0 U) H  1  a) —.3 NJ NJ —.3 I-’  H- 0 JrV ct H (0  Di  11)0  (0 l-.J cu CD a) 1 a) H ODV) a) .0. rt (0 11 H  Fiji (Dli  000000000D 0 0 0 00 0 0 0 1130  F-’ H H H H H H’ I—’ NJ NJ NJ NJ NJ NJ NJ U) U) U) C U) U) .  H N) Ui .0. 01 0) —.3 (00 H NJ 0’ a) a) a) 0) (0 0 1-’ a) NJ H (Ti H a) (Ti a) . 00) .0. 0) 0 U) 0 U) ‘ a) 0—Ja)a)(Y1(0--.Ja) U)H’ U) a) NJ —1 U)  0000 H a) —.3 a) (00 01 NJ NJ NJ U) 0 (0 a) H (0 —.3 a) —.3 0 0)  000 0000000000000  I-’  F-3 T(  Q Its  CO a) a) -J -.3 —.3 -.3 —J a) a) a) a) a) 01 (31 (.31 01 01 .0. .0. 0. 0. U) U) U) U) U) NJ NJ NJ NJ NJ H H H’ H F-’ (0 —3 01 U) I—’ h-( (15 U) H 1)1 —.3 01 Ui F—’ (0 —.3 01 U) H (0 —3 (31 Ui H (0 ——1 01 U) H (0 - 01 U) H (0 —3 01 U) H’ (0 -.3 01 U) H I)  0  DO  0000000000 NJ NJ NJ NJ NJ NJ NJ NJ NJ NJ 00 H 0 H’ H N) NJ NJ Ui .0. a) F—’ 01 —1 o Ui —3 0. (0 (0 F-’ 0. (00 0) a) a) U) a)01H’a)U)0 —JHH U)Q10)—i a) a)(ONJ  0000  H 10 10 (0 (0 (0 (0 (0 (0 (0 0 (0 (0 (0 a) a) a) a) —.3—i 0  87 89 91 93 95 97 99 101 103 105 107 109 111 113 115 117 119 121 123 125 127 129 131 133 135  0.0398668 0.0351642 0.0320662 0.0274708 0.0253071 0.0222681 0.020281 0.0191351 0.0176305 0.0163685 0.0153674 0.0142198 0.013355 0.0117696 0.0106941 0.00919034 0.00757812 0.00642323 0.00481656 0.00354516 0.00233036 0.00143848 0.000770456 0.000304338 7.59309e—005  The Primary Particles of Number for Agglomerate 1: Number of primary particles N L W 1 24.06 2563 2 29.38 27.52 3 22.51 17.20 4 15.05 14.43 5 15.00 18.13 25.95 14.38 6 7 19.39 24.38 8 36.88 26.57 9 33.13 25.00 10 22.50 20.63 11 27.19 25.98 12 28.76 26.26 13 40.03 23.44  =  13  **********  +++++++++++++++++++++++++END OF FILE+++++++++++++++++++++++++  145  APPENDIX C MATLAB CODES FOR IMAGE PROCESSING  Main Program % Main Program accepts almost all of the user input,  and manages all of  % the functions involved in the image processing dc close all fprintf(’  Start Main_Program  % =========ADMINI STRATIVE INPUT==========  % to allow user for inputting the necessary information fprintf(’Please enter the following information: session_date location grid_ID engine  input(’Location:  =  =  input(’TEM Session Date  =  input(’Grid ID: input(’Engine:  engine_mode  =  =  =  magnification  =  voltage_TEN  input  n_agglomerates file  =  ‘,  input  =  ‘s’);  ‘s’); ‘,  ‘s’); (dd/mm/yy):  ‘,  ‘a’);  ‘5’);  (‘Magnification:  (‘TEN Voltage:  input  ‘,  ‘5’);  input(’Engine Test Date  input(’Fuel:  =  ‘,  (dd/mm/yy):  ‘s’);  input(’Engine Mode:  engine_test_date fuel  ‘,  ‘,  \n\n’);  ‘,  I);  ‘a’);  (‘Number of Agglomerates in original Image:  input(’File location:  ‘,  ‘);  ‘s’);  % ==========DIGITAL IMAGE ANALYSIS========== % to process the digital images,  compute relevant information,  and  % create a text file for the database fprintf( ‘\nCommence Semi—Automatic Image Anaiysis\n’); 10  =  input(’\nPile to be analyzed in “.tif” format:  ‘,  ‘s’);  % to repeat the binary image process until a satisfactory binary image % has been obtained.  At this point in time,  the loop allows the user  146  % to change the TEM image being processed satisfaction  0;  =  while satisfaction < 1 close all;  % ---FUNCTION:  BINARY--  % this function processes the digital images to provide: % cropped image  (12),  binary image of the particle  % dilated sobel’s edge  (E4),  and super-imposed agglomerate contour  % on the original cropped image [Ii,  12,  satisfaction  BW3, =  =  14,  input(’Is  if satisfaction IC  E4,  (BWS),  (14)  F2,  this  13]  image  function_binary  =  satisfactory,  yes  (10,  [1]  or  file); no  [0]:  0  ==  input(’\nPile to be analyzed in “.tif” format:  ‘,‘s’);  end end close all  % ---FUNCTION:  CALCULATIONS--  % this function calculates the scale factor[nm per pixel]and % agglomerate’s area,  perimeter,  length,  width,  density—density  % and pair correlation function [scale factor, length_real, nprimary,  area_agglomerate_real, widthreal,  aved]  file);  % ---FUNCTION:  DIARY--  PCF,  row,  RADIUS,  function_calculations  =  magnification,  DDC,  perimeter_particle_real, col, 12,  (Ii,  % this function prints the database information into a [x]  =  function_diary  engine mode,  (session_date,  enginetestdate,  n_agglomerates,  file,  fuel,  location,  length_real,  widthreal,  PRIMARY LW,  nprimary,  DDC,  PCF,  row,  .dat  IC,  BW3,  (text)  grid_ID,  magnification,  area_agglomerate_real,  PRIMARY LW, E4,  file  engine,  voltage TM,  perimeter_particle_real, RADIUS,  col,  scale_factor,  aved);  147  fprintf  C  ‘—  End Main Program- \n\n’);  % End of Code  148  Function Binary function  [Ii,  12,  BW3,  E4,  14,  P2,  13]  =  function binary  (IC,  file)  % ===—====——=PROCESS IMAGE=========== fprintf (‘ \n\n——-Start Code function binary——--\n’); % ---FILE INPUT-adj Ii  input(’Please input the adjustment threshold level:  =  ‘);  imread(IO);  =  % ---CROPPED IMAGE 1--fprintf(’Please  crop  the  image  to  the  particle  as  close  as  possible,  cropping out the bottom white tag.’\n’) 12  imcrop(Il);  % ---FILTERED IMAGE 1--% to use the specialized “average” filter for filtering 12 h  fspecial(’average’);  =  Fl  imfilter(I2,  =  h);  % —-—FILTERED IMAGE 2-—— % to use the median filter2 function for examining a neighborhood WxW % matrix,  and making the centre of that matrix the median of the  % original neighborhood W  5; medfilt2(Fl,  P2  % figure,  [W WI);  imshow(F2),  title  (‘Filtered Image’)  % ---BINARY IMAGE 1--% to create binary image via threshold value level level2 BW  graythresh(F2);  =  =  level + adj;  im2bw(F2,level2);  =  % ---BINARY IMAGE 2,  DILATED--  % to reduce initial noise and fill initial gaps SF1  =  strel(’square’,5);  BW2  =  imdilate(—BW,SE1);  149  % figure,  imshow(BW2),  % ---BINARY IMAGE 2b,  title  (‘Binary Image 2,  Dilated’)  SELECTION--  % to filter out the smaller background objects in an image SE2  strel(’disk’,lO);  =  BW2b  =  imopen(BW2,SE2);  % ---BINARY IMAGE 3,  SELECTION--  % to further reduce the noise,  and solve the area calculation problems  % of multiple particle images fprintf(’\n\nPlease analyzed.\nDouble  select  click,  or  which right  particle(s)  click,  or  wished  shift—click on  the  to  be  desired  particle. \n’ fprintf(’Note  that  only  one  particle  may  be  selected  per  analysis! \n\n’) BW3  =  bwselect(BW2b,8);  % ---CROPPED IMAGE 2,  IMPOSED--  % to eliminate the semi—large  carbon frames in the background  % and impose the inverse of BW3 on 12 13  imimposemin(Fl,  =  % ---EDGE IMAGE 1,  BW3);  SOBEL--  % to use Sobel’s Method to create particle’s outline El  edge(13, ‘sobel’);  =  % ---EDGE IMAGE 2,  DILATED--  % to strengthen the particle’s outline, SE2 E2  =  =  use dilation  strel(’square’,2); imdilate(E1,SE2);  % ---EDGE IMAGE 3,  MANUAL SELECT & REMOVE--  % to get rid of large spots that are not part of the image fprintf(’Piease the outline. E3  =  RIGHT  click on the pixels  that  are  clearly not part  of  LEFT click on the final spot you would like to remove.’);  bwselect(E2,4);  150  E4  =  E2  E3;  -  close;  % ---CROPPED IMAGE 3,  IMPOSED---  % to impose final E4 onto 12 14  =  imimposemin(12,  E4);  % -——Saving Images filename path  =  strcat(file,  figure,  imshow(12),  saveas (gcf,  filename path  =  =  filename =  =  filename path  =  =  filename path  =  =  fprintf  (  title  (‘Binary Image 3,  Selected’)  filename); title  (‘Edge Image 3,  Manually Selected & Removed’)  ‘Final Imposed Image.jpeg’;  imshow(14),  saveas (gcf,  filename);  path);  strcat(file,  figure,  (‘Filtered Image’)  ‘Edge Image.jpeg’;  irnshow(E4),  saveas (gcf,  title  path);  strcat(file,  figure,  filename);  ‘Binary Image Imagejpeg’;  imshow(BW3),  saveas (gcf,  (‘Cropped Image’);  path);  strcat(file,  figure,  title  ‘Filtered Image.jpeg’;  imshow(F2),  saveas (gcf,  filename);  path);  strcat(file,  figure,  path  ‘Cropped Image.jpeg’;  filename); title  (‘Cropped Image 3,  Final Imposed Image’)  path);  ‘\n\n---End Code function_binary---\n’);  151  Function Calculations function  [scale_factor,  length_real, nprimary,  widthreal, aved]  magnification,  area_agglomerate_real, DDC,  PCF,  row,  RADIUS,  function_calculations  =  perimeter_real, col,  (Ii,  12,  PRIMARY LW, BW3,  E4,  file)  % this function was developed to extract geometrical properties of the % soot aggregate  fprintf(’ \n\n—-—Start Code function calculations---\n’);  % =========CALCULATIONS========== % to find image Y figure,  handle  =  height and X  =  =  width  imshow(BWS);  imgmodel=imagemodel (handle); Y  =  getlmageHeight(imgmodel);  X  =  getlmageWidth(imgmodel);  close;  % ---FUNCTION:  SCALE---  % to determine the scale factor [scale_factor]  % ---FUNCTION:  =  [nm per pixel]  function scale3  (Ii, magnification);  PRIMARY PARTICLE LENGTH AND WIDTH--  % to determine the size of the primary particles,  and their  % individual lengths and widths [PRIMARY LW, scale_factor,  % ---FUNCTION:  nprirnary,  aved]  function_primary_particles (12,  =  file)  AREA TOTAL--  % to determine the total projected area of the aggregate [area_agglomerate]  =  area_agglomerate_real  % ---FUNCTION:  function_area_total =  area_agglomerate  (BW3); *  scale factor”2;  PERIMETER--  % to determine the perimeter of the aggregate [perimeter]  =  function_perimeter3  (2W3);  152  *  perimeter real  =  % ---FUNCTION:  LENGTH AND WIDTH--  perimeter  scale_factor;  % to determine the length and width of the aggregate width]  [length,  length_real width_real  function length width3  =  length  =  *  width  =  *  (54,  file);  scale_factor;  scale_factor;  SLOPE--  % ---FUNCTION:  % to determine the relationship between the threshold value and % the radius [trcoefficient]  function tr coefficient  =  (X,  Y);  % ---FUNCTION: AREA R WHOLE-% to calculate ideal case for r  —  (r—l),  for the density density  % correlation [AREA RWHOLE]  =  % ---FUNCTION:  function_area_r_whole  (trcoefficient,  X,  Y);  DENSITY DENSITY CORRELATION--  % to determine the density density correlation % this function provides DDC in an array: % column 1  =  increasing r[nm]  % column 2  =  the value C(r)  [DDC,  row]  =  tr_coefficient,  % ---FUNCTION:  function density density correlation4 areaagglomerate, AREA RWHOLE,  file,  (BW3,  X,  Y,  scale_factor);  PAIR CORRELATION FUNCTION--  % to determine the pair correlation function RADIUS,  [PCF,  col]  =  functionpair correlation function2  (BW3,  file,  scale_factor);  % ---DDC & PCF FIGURE-% to directly compare DDC and PCF on the same graph filename  =  ‘Density  Density  Correlation  and  Pair  Correlation  Function  Graphed Together.jpeg’; path  =  strcat(fiie,  filename);  153  figure,  plot(  DDC(:, 1),  xlabel(’radius’), Function and  DDC(: ,2),  ylabel(’PCF  and  ‘—b’,  DDC’),  RADIUS’,  title(’The  Density-Density Correlation Graphs’),  PCF’, Pair  ‘—r’),  Correlation  legend(’DDC’,  ‘PCF’,  —1) saveas (gcf,  fprintf  (  path);  ‘\n\n———End Code function calculations--—\n’);  % End of function_calculations  154  Function Scales function  [scale_factor]  =  function scaie3  (Ii, magnification)  % this function determines the conversion factor between pixel and real % length from the black scale bar on the TEM image  % to crop the image and obtain the scale bar fprintf(’Please crop to the image’’s scale.\n’); S  =  imcrop(Il);  % to get X, figure,  Y axis of the original image  handle  (Il)  imshow(S);  =  imgmodel=imagemodel (handle); Y  =  X  getlxnageHeight(imgmodel); getlmageWidth(imgmodel);  close;  % to find the longest line of black pixels in S count  =  0;  largest_count  for y  =  0;  l:l:Y  for x  =  l:l:X  if S(y,x) count  ==  =  if count  0 count >=  +  1;  largest count  largest_count  =  count;  end end end count  =  0;  end % to calculate the scale factor scale_factor  =  magnification  /  largest count;  % End of function scale3  155  Function Area_Total function  [area agglomerate]  =  function area total  (3W3)  % this function finds the total area of the agglomerate by counting all % of the nonzero pixels  area agglomerate  =  nnz  (BW3);  % End of function_area_total  156  Function Perimeter function  [perimeter]  function perimeter3  =  (BW3)  % this function finds the perimeter of the agglomerate by counting the % white pixels that are adjacent to the black pixels  % to initialize variables and locate all white pixel indices  [ROW, COLI perimeter  =  =  find  (BW3);  0;  area_agglomerate  nnz(BW3);  =  % accumulating the perimeter count for pos if  =  l:l:area agglomerate  (BW3(ROW(pos)—1, perimeter  =  COL(pos))  ==  0)  ==  0)  ==  0)  ==  0)  perimeter + 1;  end if  (Bw3(ROW(pos), perimeter  =  COL(pos)+l)  perimeter + 1;  end if  (8W3(ROW(pos), perimeter  =  COL(pos)—l)  perimeter + 1;  end if  (BW3(ROW(pos)+l, perimeter  =  COL(pos))  perimeter + 1;  end end  % End function_perimeter  157  Function Length_Width function  [length, width]  =  function length width3  (E4,  file)  % this function determines the length & width of the agglomerate,  and  % provides a rotated image of the particle with length and width axis  %  to  initialize  variables  and  ‘find’  the  white  pixel  indices  of  the  edge % image [ROW1,  COL1] = find  (E4);  area_edge_particle = nnz(E4); length = 0;  % ----LENGTH-% to compute the length in vector form, % two pixels that specify the length. % search column—by--column, % increasing y,  and record the position of the Since the ‘find’  function es  in the order of increasing x,  then  the pixel specified by ‘posi’ will always have a  % greater y—indices than the pixel specified by pos2 for posi = i:i:area edge particle DISTANCE  (  (001,1  —  COL1(posl)).A2  [greatestdistance, pos2]  +  (ROW1  —  ROW1(posi))/’2 ).“.5;  max(DISTANCE);  if greatest_distance >= length length = greatest_distance; xlibot = COL1(posl); yllbot = ROW1(posl); x2_l_top = COL1(pos2); y2ltop  =  ROW1(pos2);  end end  % to mark the pixels that specify the length.  It is important to note  % that an error may arise during the rotation of the image,  because  % some pixels are obscured,  or even deleted via the rotation process.  % In order to prevent this,  the pixel left—adjacent to the true length  % defining pixels are also marked with a value of ‘2’,  or ‘3’  E5 = E4;  158  E5  (yllbot,  xllbot: (xllbot+2))  E5  (ylibot,  xllbot+1)  E5  (y2ltop,  x2ltop)  E5  (y2ltop,  x2ltop+1)  =  2;  2;  =  3; =  3;  % ———WIDTH-—— % to determine theta,  the problem becomes a system of 2 equations with  % 2 unknown, with 2 differing situations if xllbot theta  x2_l_top  ==  0;  =  end if xllbot < x2ltop theta  180  =  abs(atan(  (x2ltop  -  xllbot)  /  (ylibot  y2ltop)  —  )  *  / pi);  end if xllbot theta  x2ltop  >  abs  =  y2ltop)  (180 *  )  180  (atan(  —  (xllbot  —  x2ltop)  /  (ylibot  —  / pi) );  end  % to rotate the image counterclockwise,  until the length axis is  % vertical ROT  =  imrotate  (E5,  theta);  % to find the maximum and minimum x values to calculate the width i  =  0;  [ROW2,  COL2]  find  =  [xlwrit,  i]  ylwrit  ROW2 (i);  =  =  max(COL2);  [x2wlef,  ii  y2wlef  ROW2(i);  width  =  =  (ROT > 0);  rnin(COL2);  =  xlwrit  -  x2wlef;  % ----PLOT LENGTH AND WIDTH AXIS ALONG ROT-[ROW3,  COL3]  =  find  (ROT  ==  2);  159  if ROT  (ROW3(1),  COL3(1))  ==  xllbot rot  =  COL3(l);  ylibot rot  =  ROW3(l);  2  end  [ROW4,  COL4]  if ROT  (ROW4 (1),  =  find  (ROT  ==  COL4 (1))  ==  x2ltop rot  =  COL4(1);  y2_l_top_rot  =  ROW4 (1);  3); 3  end  % to plot and save ROT with length and width,  where the length is the  % blue vertical line and the width is the red horizontal line.  Note  % that they are perpendicular to each other filename path figure,  =  ‘Length and Width Image.jpeg’;  strcat(file, imshow(ROT),  filename); title(’ROT,  Length  (Blue)  & Width  (Red)  Axes’)  hold on line  ([xl. 1 bot rot,  ‘linewidth’, line  line  saveas (gcf,  y2 1 top rot],  x2wlef],  [y2wlef,  y2wlef], ‘Color’,  ‘r’,  xlwrit],  [ylwrit,  y2wlef],’Color’,  ‘r’,  3);  ([xlwrit,  ‘Linewidth’,  [yl 1 bot rot,  3);  ([xl_w_rit,  ‘linewidth’,  x2 1 top rot],  3); path);  % End of function length width  160  Function TR_Coefficient function  [trcoefficient]  =  function tr coefficient  (X,  Y)  % this function determines the tr_coefficient of the relationship % between threshold value and the radius of the circle,  thus allowing  % one to control the radius of a circle by controlling the threshold % value  % to place a white pixel in the middle of an image that is identical in % height and width to the image being processed BW4  zeros(Y,  =  X);  yl  =  uintl6(Y  /  2);  xl  =  uintl6(X  /  2);  xl)  1;  BW4(yl,  =  % to create the distance transform of BW4 BW5  =  bwdist(BW4,  ‘euclidean’);  % to create a pseudo-grayscale image of BW5 for col  l:l:Y  =  MAX(col)  max(5W5(col,:));  =  end maximum BW6  =  max(MAX);  =  ./ maximum;  BW5  % to calculate and record the increasing radius,  given the increasing  % threshold value THRESH  0;  =  R1D  =  0;  row  =  0;  for thresh  =  0.001:0.001:.5  row  =  row  BW7  =  im2bw(5W6,  area r  =  =  +  1; thresh);  nnz(BW7);  sqrt  (area/pU;  THRESH(row)  =  thresh;  161  RAD(row)  =  r;  end  % to calculate the trcoefficient row  =  row  —  1;  trcoefficient  =  (RAD(row)  —  RAD(2))  /  (THRESH(row)  —  TJ-IRESH(2));  % End of function tr coefficient  162  Function AreaR_Whole function  [AREA RWHOLE]  function arearwhole  =  (tr_coefficient, X,  Y)  % this function calculates the area of a black ring of an outer radius % r,  and thickness of dr  =  1,  and returns it in matrix form,  with  % increasing r.  % to initialize the variable row  =  0;  % to create a the biggest possible distance transform  (one corner to  % the other) BW1  =  yl xl  zeros(Y, X); uintl6(Y/2); uintl6(X/2);  =  BW1(yl,xl) BW2  =  1;  =  bwdist(BWl);  % to find the maximum distance in 3W2 for col  =  l:l:Y  M1X(col)  max(BW2(coi,:));  =  end maximum  =  max(MAX);  % to create the largest pseudo—grayscale image which encompasses the % largest possible r X2  =  uintl6(4  *  maximum(l));  Y2  =  uintl6(4  *  maxiiuum(l));  BW3  =  zeros(Y2, X2);  y2  =  uintl6(Y2  /  2);  x2  =  uintl6(X2  /  2);  BW3(y2, x2)  =  1;  BW4  =  bwdist(BW3);  3W5  =  BW4  ./  maximum;  % to create threshold proportional to radius,  create the black circle  163  % and calculate/store the data for r  l:1:tr coefficient  row  row + 1;  =  thresh BW6  =  =  r  /  tr_coefficient;  im2bw(BW5,  AREA CIRCLE(row,  thresh); 1)  AREA CIRCLE(row,  r; 2)  1  =  function_area_total  (BW6);  end  % to calculate AREA RWHOLE AREA RWHOLE(1:row, AREARWHOLE(1,  2)  AREA RWJ4QLE(2:row,  1) =  =  AREA CIRCLE(:,  AREA CIRCLE(1, 2)  =  2)  1); -  1;  AREA CIRCLE(2:row,  2)  -  AREA CIRCLE(l: (row-l),  2);  % End of function area rwhole  Function Density_DensityCorrelation function  [DDC,  tr_coefficient,  row]  =  function density density correlation4  areaagglomerate,  AREA RWHOLE,  file,  (BW3,  X,  Y,  scale_factor)  % this function determines the density density correlation function  % to initialize variables DDC  0;  =  AREAR  0;  AREAR(1,  2)  density  200;  =  =  0;  % to find the maximum value given the height and width of the image,  in  % which to create pseudo—grayscale images 3W4  =  zeros(Y,  X);  yl  =  uintl6(Y/2);  xl  =  uintl6(X/2);  BW4(yl,xl) BW5  =  =  1;  bwdist(BW4);  164  for col  =  i:l:Y  MAX(col)  max(BW5(col,:));  =  end maximum  max(MAX);  % to find the indices of all white pixels, [ROW,  COL]  find  =  representing the aggregate  (8W3);  % to determine AREA R -the number of pixels of radius r % that reside within the boundary of the aggregate for k  =  BW4  i:density:area agglomerate =  zeros(Y,  BW4(ROW(k),  X);  COL(k))  BW5  =  bwdist(BW4,  BW6  =  8W5  .1  maximum;  BW7  =  3W3  •k  BW6;  row  =  0;  for r  =  row  ‘euclidean’);  l:1:tr coefficient =  thresh BW8  1;  =  =  row + 1; =  r  /  im2bw  tr_coefficient; (BW7,  [area_remainder] area_r  =  thresh); function_area_total  =  area_agglomerate  (BW8);  area_remainder;  —  AREA R(row,  1)  =  r;  AREA_R(row,  2)  =  AREA_R(row,  2)  +  area_r;  end end  % to find AREAR PARTICLE, % thickness dr  =  1,  the number of pixels of outer radius r,  and  that reside within the boundary of the  % aggregate AREAR PARTICLE (1, 1) =0; AREARPARTICLE(1,2)  =  AREARPARTICLE(2:row,  % to find DDC,  AREAR(l) 2)  =  -  1;  AREA_R(2:row,  2)  -  AREA R(l: (row-i),  2);  the density density correlation  165  DDC(l:row,  1)  =  AREA R(1:row,  DDC(1:row,  2)  =  AREARPARTICLE(l:row,  DDC(l:row,  2)  DDC(1:row,  1)  •*  scale_factor; 2)  ./ AREA RWHOLE(l:row, 2);  ./ (area agglomerate/density);  2)  % to calculate the density density correlation.  It is defined that the  % value of C(r)to be equal to the difference in areas between the sum % of the areas of r and the sum of the areas of  (r—l),  and then the  % whole thing divided by the number of pixels considered filename path  =  =  ‘Density Density Correlation F’unction.jpeg’;  strcat(file,  figure,  plot  filename);  (DDC(l:row,  1),  DDC(l:row,  2)),  xlabel  (‘radius’),  ylabel(’C(r)’); saveas (gcf,  path);  % to create the logarithmic plot filename path  =  figure,  =  ‘Density Density Correlation Function.jpeg’;  strcat(file, loglog  filename);  (DDC(1:row,  1),  DDC(l:row,  2)),  xlabel  (‘radius’),  ylabel(’C(r) ‘); saveas (gcf,  path);  % End of function density density correlation4  166  Function Pair_Correlation unction file,  {PCF,  RADIUS,  col]  =  function pair correlation function2  (BW3,  scale factor)  % this funcrion determines the pair correlation function and provides % the plot of the pair correlation function versus radius  % to initialize all variables,  and ‘find’  the indices to all the white  % pixels in BW3 DISTANCE ALL DEN0MINATORR X  =  0;  Y  =  0;  0;  =  0;  =  dr =2; m  =  0;  density [ROW,  30;  =  COL]  find  =  area_particle  j  =  (BW3);  nnz(BW3);  =  0;  % to consolidate the pixels of consideration to much smaller arrays for k  l:density:area particle  j  j  1;  +  X(j)  =  COL(k);  Y(j)  =  ROW(k);  end  % to calculate all the distances with reference to one pixel at a time, % using vector algebra, N  and then adding the results  =  for  j  =  l:1:N  DISTANCE  =  (  (X  DISTANCE ALL(  —  ((j  X(j)).”2 + -  l)*N+1)  (Y :  —  Y(j)).”2  (j*N)  )  =  ).“.5;  DISTANCE;  end  % to create the  ‘RADIUS’  bins,  in preparation for when the  ‘PCF’  array  % is put into a histogram  167  greatest_distance nbins  greatest_distance;  =  DISTANCE ALL nbins  nonzeros (DISTANCE ALL)  =  nbins  =  RADIUS  double (uintl6 (max(DISTANCE ALL)));  *  •*  scale_factor;  scale_factor;  1:dr:nbins;  =  % to save the distribution of distances as the distribution histogram filename path  ‘Pair Correlation Histagram.jpeg’;  =  strcat(file,  =  figure,  filename);  hist(DISTANCE ALL,  RADIUS),  title  (‘Pair Correlation  Histogram’) saveas (gcf,  path);  % to create  ‘PCF’,  % according to PCF  the frequency of distances in  ‘RADIUS’  hist(DISTANCE ALL,  =  % to create the  ‘DISTANCE_ALL’  RADIUS);  ‘DENOMIATORR’  and normalize the pair correlation  % function for r  =  l:dr:nbins  m  =  m + 1;  DENOMINATORR(m)  =  (2*pi*N*r*dr  / (scale factor2*density));  end  ./ DENOMINATORR;  PCF  PCF  col  length(PCF);  % to save the pair correlation function in histogram format filename path  =  =  ‘Pair Correlation Function Histogram.jpeg’;  strcat(file,  figure,  bar(RADIUS,  filename); PCF),  (‘radius’),  ylabel(’PCF’)  saveas (gcf,  path);  title  (‘Pair  Correlation  Histogram’),  % to save the pair correlation function in line graph format filename path  =  =  ‘Pair Correlation Function Plot.jpeg’;  strcat(file,  filename);  xlabel  figure, xlabel  plot(RADIUS, (‘radius’),  saveas (gcf,  PCF,  ‘—r’),  title  (‘Pair  Correlation  Line  Plot’),  ylabel(’PCE(r)’)  path);  % End of function pair correlation function2  169  Function Primary_Particles function  [PRIMARY LW,  scale factor,  nprimary,  aved]  =  function primary particles(I2,  file)  % this function applies the user’s selected input to identify the % number of primary particles, % diameters.  and their individual lengths,  widths and  After estimating the number of primary particles within  % the particle,  the user selects the endpoints of the length,  % diameter respecitvely,  for each primary particle.  width and  The length has  % been defined as the major axis parallel to the axis of the length of % the aggregate that encompasses the greatest distance that the primary % particle spans.  The width has been defined as the minor axis,  % is the greatest distance that the primary particle spans,  which  that is  % perpendicular to the major axis.  % to upload the image to Matlab image  12;  =  figure, Width  imshow(image,  (Red)  [1),  title(’Primary  Particles,  Length  (Blue)  &  Axes’);  % to input the number of the observed primary particles nprimary  =  length  zeros(nprimary,l);  width  =  zeros  =  totald  input(’Please input the number of primary particles:  =  ‘);  (nprimary,l);  0;  % to input the points which correspond to the number % of the primary particles using the cursor for index  =  l:l:nprimary  [x,  yl  =  ginput(2);  x  =  x;  y  =  y;  % Calculating the length of each primary particle by using eucledian % distance formula  length(index,1)  =  scale factor*sqrt((x(2)_x(l))/2+(y(2)  —  % to draw blue lines that represent the length of each primary  170  % particle hold on line  [a, b] a  =  b  =  =  ([x(l),x(2)],[y(l),y(2)],  ‘linewidth’,  3);  ginput(2);  a;  % to calculate the width of each primary particle by using eucledian % distance formula width(index,1)=scalefactor*sqrt((a(2)_a(1))A2+(b(2)  —  % to draw red lines that represent the width of each primary % particle hold on line  ([a(1),a(2)],[b(1),b(2)],’Color’,  [c,  dl  ginput(2);  c  =  C;  d  =  d;  =  ‘r’,  ‘linewidth’,  3);  % to calculate the diameter of each primary particle using % eucledian distance formula dia(index, l)=scale factor*sqrt( (c(2)—c(l) )2+(d(2) totald  =  totald  +  —  d(1) )2);  dia(index,l);  end  % to calculate the average diameter of the primary particles aved  =  totald  filename path  =  /  nprimary;  ‘Primary Particle Length and Width Image.jpeg’;  =  strcat(file,  filename);  saveas (gcf, path);  numindex=  (l:1:nprimary);  PRIMARY LW  =  [numindex’,  length, width];  % End of function primary particles  171  Function Diary function  [xl  function diary  =  engine mode,  (session_date,  engine_test_date,  n_agglomerates,  file,  fuel,  location,  magnification,  area_agglomerate_real,  lengthreal,  widthreal,  PRIMARY LW,  nprimary,  DDC,  PCF,  row,  grid_ID,  10,  engine,  voltage TEM,  perimeter_particle_real,  RADIUS,  col,  scale_factor,  aved)  % this function prints the information on the aggregates into the text %  x  (dat*)  =  file using the default diary function.  1;  % to begin printing into the text file. path diary  strcat(file,  =  ‘Database Eile.dat’);  (path);  (  fprintf  ‘\n  START  PILE  fprintf( ‘\n==============ADMINISTRATION==============\n’);  fprintf  (  ‘  fprintf(’TEM Session Date  (dd/mm/yy):  fprintf( ‘Location:  location);  fprintf(’Grid ID: fprintf(’Engine:  %s\n’, %s\n’,  %s\n’,  fprintf(’Engine Mode: fprintf(’Engine Test  fprintf(’Fuel:  %s\n’,  fprintf(’Number  of  session_date);  grid_ID); engine);  %s\n’, Date  engine_mode);  (dd/rnrn/yy):  %s\ri’, engine_test_date);  fuel);  fprintf(’Magnification: fprintf(’TEM Voltage:  %s\n’,  %g\n’,  %s\n’,  magnification);  voltage TEM);  Agglomerates  in  original  Image:  n_agglomerates); fprintf(’Image Name:  fprintf(’  fprintf  (  %s\n’,  10);  \n’);  ‘\n\n=================ANALYSIS===============\n’);  % to create a loop for analyzing image with more than 1 agglomerate  172  while n_agglomerates > 1 fprintf(’The Agglomerate of Number:  fprintf  (  %.Of\n’,  x);  ‘  fprintf(’Scale Factor fprintf(’Length fprintf(’Width  [nmj  [nm/pixel] %g\n’,  =  [nm]  %g\n’,  fprintf(’Perimeter  %g\n’,  fprintf(’Area  [nm”2]  =  scale_factor);  width_real);  [nm]  =  %g\n’,  length_real);  fprintf(’Average Diameter [nm]  =  =  %g\n\n’,  %g\n’,  aved);  perimeter_particle_real); area_agglomerate_real);  fprintf (‘Density-Density Correlation:  \n’);  fprintf(’There are %.Of data points.\n’,  row);  fprintf( ‘r\tC(r) \n’) for  j  l:l:row  =  fprintf(’%.Of\t%g\n’,  DDC(j,  1),  DDC(j,  2));  end fprintf(’\nPair Correlation Function:  \n’);  fprintf(’There are %.Of data polnts.\n’, fprintf  col);  ( ‘rad\tC (rad) \n’);  j  for  l:1:col  =  fprintf(’%.Of\t%g\n’,  RADIUS(j),  PCF(jH;  end fprintf( ‘\n  \n’);  fprintf(’\nThe %.Of:\n’,  Primary  Particles  1;  j  =  1;  fprintf for k  =  j+1), =  Agglomerate  =  \t%g’,  nprimary);  ( ‘\nN\tL\tW\n’); 1:1:nprimary  fprintf  i  for  \nH;  fprintf(’Number of primary particles =  Number  x);  fprintf(’  i  of  (  ‘%  .  Of\t% 2f\t% 2f\n’, -  PRIMARY LW(i,  j+2)  PRIMARY LW(i,  j),  PRIMARY LW(1,  )  i+1;  173  end fpr±ntf(  fprintf  (  —  !**********\fl\flI)  x+l;  x  n_agglomerates  =  n_agglomerates  -  1;  end  fprintf  (  \n+++++±±+++++++÷±++++++++±END  OF  diary off  % End function_diary  174  Read Program % This program reads the Batch file that specifies which database text %  files should be read.  close all  fprintf  (  Start  ‘  Read_Program  \n’);  % =========OBTAIN USER INPUT==========  % to determine where to store the “batch results” file file ,  ,  input(’Piease  =  enter  the  file  location  to  store  the  results:  ,  ,  ,  % to find the “batch” file text_file  input(’Please  database files:  ‘,  enter  the  exact  pathway  for  the  textfile  of  ‘a’);  % to scan the batchfile for the database text files to be read fid=fopen(text file); data_files  textscan(fid,  =  ‘headerLines’, data_files n_files  =  =  ‘delimiter’,  ‘%s’,  ‘whitespace’,  ‘\n’,  ‘  4);  data files{:};  length(data files);  % ---FUNCTION:  FILE READER--  % this function scans the files using the  ‘textscan’  function,  and  % returns the read data in “cell” format [session_date,  location,  engine_test_date,  fuel,  filename, perimeter,  NUN AGM, area,  (data_files,  %  %  magnification,  scale_factor, DDC,  PCF,  engine,  engine_mode,  voltage TEM,  n_agglomerates,  length_agglomerate,  DDC_R,  PCFR,  x,  y]  =  width_agglomerate,  function_file_reader  n_files);  ---FUNCTION:  this  grid_ID,  function  FILE PRINTER--  prints  the  results  of  all  the  database  text  files,  where  175  % each column represents the results from one database text file [x]  =  function_file_printer  engine_mode,  engine_test_date,  n_agglomerates,  filename,  width_agglomerate, nfiles,  fprintf  (  x,  (session_date, fuel,  NUMAGM,  perimeter,  area,  location,  grid_ID,  magnification,  scale_factor, DDC,  DDCR,  engine,  voltage TEM,  length_agglomerate, PCFR,  PCF,  file,  y);  ‘  End  Main Program  % End of Code  176  APPENDIX D MATLAB CODES FOR FRACTAL ANALYSIS  Fractal Analysis Program (LW Method) % This program automatically calculates the fractal dimension based on % aggregate projected dimension method as well as its 95% confidence % interval  dc close all fprintf(’\n\n-—-Start Code Fractal Analysis based on  rep  input(’The number of files that will be analyzed:  =  dloc  =  (.dat*):  input(’Diary location  ‘,  ‘  );  ‘s’);  for index =l:1:rep  % ---FILE INPUT---  file  =  input(’File location:  dmarker  input(’line  =  (+,o,*, .,s,d,p,h) rmarker I  I  fuel  =  ,  ‘s’);  color  (r,g,b,c,m,y,k..w)and  for the data plot e.g vgd:  input(’line  color  ,  marker  type  st);  (r,g,b,c,m,y,k,w)for  the  regression  line:  I  =  engmode  input( f 1 uel type: =  irlput(Tengine  I,  ‘s’);  mode:  I, ‘s’);  diary (dioc); fprintf(’Fuel type:  %s,  Engine mode:  , 1 %s\n  fuel,  % to read the data from the spreadsheet file, % column of the data to the vector N  engmode)  and assign the first  (number of primary particles)  % and the second column to vector A (projected area of the aggregate) data  =  xlsread(file);  177  A  data(:,2);  =  Asmall  =  Abig  max  =  mm  (A); (A);  gap  =  N  data(:,l);  =  fix((Abig—Asmall))/30;  % to fit the data using the power function p  polyfit(log(A),log(N),l);  =  xt  =  yt  loglO (A); loglO(N);  [fresult,gof]  =  fit (xt,yt, ‘polyl’)  % to plot the scattered data and its curve—fitted line on a logarithmic % scale Abot  =  Asmail_4*gap;  Atop  =  Abig + 4*gap;  Ax  Abot:gap:Atop;  =  lnNy  =  Ny  exp  =  polyval(p,  log(AxH;  (lnNy);  loglog(A’ ,N’ ,dmarker,Ax,Ny,rmarker) xlabel( ‘LW’{2}/dp  [m’{2}] ‘);  ylabel(’N’);  hold on diary off; end hold off  fprintf(’\n\n--—End Code Fractal Analysis based on LW—--\n’);  % End of Code  178  Fractal Analysis Program (PCM Method) % This program automatically calculates the fractal dimension based % on pair correlation methods as well as its 95% confidence % interval  dc close all fprintf(’\n\n--—Start Code Fractal Analysis based on  rep  input(’The number of files that will be analyzed:  =  dloc  (.dat*) :  input(Diary location  =  T,  );  ‘s’)  for index =l:l:rep  % ---FILE INPUT--file  input(’File location:  range  input(’Range  =  (i.e: Al:B4’, dmarker  V  V  fuel  data  input(’line  =  used  for  fractal  analysis  based  on  PCM:  ‘5’);  (+,o,*, .,s,d,p,h) rmarker  of  ,  color  (r,g,.b,c,m,y,k,w)and  for the data plot e.g “gd”:  input(’line  color  ‘,  marker  type  ‘s’);  (r,g,b,c,m,y,k,w)for  the  regression  line:  V  =  engmode  iriput(’fuel type: =  ‘,  ‘s’);  input(’engine mode:  ‘,  ‘s’);  diary (dloc); fprintf(’Euel type:  %s,  Engine mode:  %s\n’,  % To read all data from the spreadsheetfile, % column of the data to the vector Cfull  % and the second column to Rfuli data  fuel,  engmode)  and assign the first  (pair correlation function)  (radius)  xlsread(file);  =  Rfull  =  data(:,l);  Cfull  =  data(:,2);  % to read the chosen data from the spreadsheetfile, % first column of the data to the  vector C  and assign the  (pair correlation function)  179  % and the second column to R PCM  =  R  PCM(:,1);  =  xlsread(file,  Rsmall  =  Rbig  max  =  miri  range);  CR);  CR);  gap  =  C  PCM(:,2);  =  (radius)  fix((Rbig—Rsmall))/30;  % to fit the data using the power function p  polyfit(log(R) ,log(C) .1);  =  xt  =  loglO(R);  yt  =  loglO(C);  [fresult,gof]  =  fit (xt,yt, ‘polyl’)  % to plot the scattered data and its curve—fitted line on a logarithmic % scale Rbot  =  Rsmall_4*gap;  Rtop  =  Rbig + 4*gap;  Rx  Rbot:gap:Rtop;  =  lnCy  =  Cy  exp  =  polyval(p,  log(Rx));  (lnCy);  loglog (Rfull’  ,  xlabel(’R’);  ylabel(’C(R) ‘);  Cfull’  ,  dmarker, Rx, Cy, rmarker)  hold on diary off; end hold off  fprintf(’\n\n——-End Code Fractal Analysis based on  % End of Code  180  APPENDIX E RAMAN SPECTROSCOPY CALIBRATION  The preliminary measurement of first-order Raman of soot showed the presence of contaminant peaks in the Raman spectrum over the Raman shift range of 500  —  2000 cm”  as illustrated in Figure E. 1. It was suspected that these contaminants peaks were caused by the defect on the grating refracting light due to the damage in the groves of the grating.  5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0  500  700  900  1100  1300  1500  1700  1900  Rann Shift (cm ) 4  Figure E.1 Contaminant Peaks in the Recorded Raman Spectra  In order to identify these contaminants peaks so that they could be excluded from the Raman spectral analysis, the Raman spectroscopy was calibrated against a pure gold  sample utilizing the fact that gold should not exhibit any first-order Raman bands. Figure  181  E.2 shows the first-order Raman spectra of gold over the Raman shift range of 500  —  2000 cm’,  Reman Spectra of Gold 500  1  450 400 350 300 .2 250 200 150 100 50 0  1000  500  1500  2000  Raman shift (cm) 1  Figure E.2 Raman Spectra of Gold  It was observed that there were four main contaminants peaks present in the Raman spectroscopy system. Table E. I lists the contaminant bands and their positions.  Table Li Positions of the Contaminant Bands Band  Position (em ) 1  Contaminant 1  900  Contaminant 2  979  Contaminant 3  1230  Contaminant4  1338  182  In order to determine their shapes, the chi-square goodness-of-fit tests were performed. The ‘Contaminant 1’ and ‘Contaminant 4’ bands both exhibited narrow bands and hence, Lorentzian line shapes were assumed for both bands. Four different combinations of contaminant bands tested are listed in Table E.2. The goodness-of-fit achieved with different band combinations is indicated by the reduced x 2 value  Table E.2 Band Combinations Tested (Line shape: G = Gaussian, L = Lorentzian) Band  Initial Position (cm’)  (I)  (II)  (III)  (IV)  Contaminant 1  900  L  L  L  L  Contaminant 2  979  L  G  L  G  Contaminant 3  1230  L  L  G  G  Contaminant 4  1338  L  L  L  L  Averaging over ten spectra yields 1.83 ± 0.3 for band combination (I), 1.24 ± 0.2 for band combination (II), 2.16 ± 0.5 for band combination (III), and 2.08 ± 0.5 for band combination (IV). This result implies that the contaminant spectra were best fitted with a combination of three Lorentzian-shaped bands (‘Contaminant 1’, ‘Contaminant 3’, and  ‘Contaminant 4’) and a Gaussian-shaped band (‘Contaminant 2’).  183  APPENDIX F SPECTRAL PARAMETERS OF SOOTS  The spectral parameters of the first-order Raman bands of different soot samples collected from engines are summarized below. These includes the band position (Stokes Raman shift 1 cni ) , full width at half maximum (FWHM,), and peak intensity ratios  relative to the G band (mean values ± standard deviation).  Table F.1 Spectral Parameters for ULSD Soot (without Oxidation Catalyst) Band  Parameter  G  Dl  D2  D3  Mode I  Mode 2  Mode 3  Position  1578.5±5.80  1578.2±4.0  1577.9±5.7  FWHM  72.21 ±6.65  73.14±6.28  73.12 ±6.13  Position  1353.9 ± 2.8  1354.5 ± 2.7  1354.2 ± 3.2  FWHM  180.05 ± 13.72  229.07 ± 11.02  275.05 ± 15.28  G 1 ‘D1’  4.32 ± 0.90  4.58 ± 0.86  4.93 ± 0.94  Position  1597.9 ± 5.8  1602.0 ± 6.0  1598.4 *5.5  FWHM  67.24± 13.72  69.51 ± 13.73  67.38± 13.74  ‘02”G  0.55 ± 0.14  0.51 ± 0.12  0.58 ± 0.16  Position  1505.2 ± 8.2  1506.2 ± 7.4  1505.4 ± 8.3  FWHM  91.09 ± 5.52  90.78 ± 7.28  92.04 ± 6.74  0.37 ± 0.13  0.38 ± 0.15  0.42 ± 0.16  Position  1186.8± 18.5  1185.7±20.7  1185.2± 19.6  FWHM  22557±68.0  242.22±65.2  244.53±61.5  ‘D3”G  D4  184  00  0  C  r,)  7J)  00  -  H-  0  1+  0 1+  1+ 0  1+  H-  -  .  P4  Cn4  1+  -  pc)  p..  -.  Z 0 Z  11O5fl  ..  D  C.) 1+ 0  a) r\) 1+  0 C.) 1+  -  CO 0) 1+  -  c  1+ 0  -  4 1+ 0  )  0) 1+  I’.3  2:  CD H  -  1+  C.)  0) H-  -.J  0 H-  C.)  -  -  (Ji  -  -  0  0 f  -  0  C.)  C..)  ()1  CD 0 Co( Q Ø 4 )  -  0) 1+  0)  -  a) 1+  0)  I 0 Z  <  0  0  CD  ii;. —‘  D 0)  a  0)  1+ 0  -  1+  -  CD H-  H0  H-  ---.j  1+  -  Co H-  1+  C)1biJ 0 0(0C0c  C) C.) 1+  1+ 0  r  0) 0 H-  D  0  -  I  S S S.  S  :  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0070805/manifest

Comment

Related Items