UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mixture formation in a partially stratified directly injected natural gas engine Shield, Malcolm 2011

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
ubc_2011_fall_shield_malcolm.pdf [ 58.17MB ]
Metadata
JSON: 1.0072272.json
JSON-LD: 1.0072272+ld.json
RDF/XML (Pretty): 1.0072272.xml
RDF/JSON: 1.0072272+rdf.json
Turtle: 1.0072272+rdf-turtle.txt
N-Triples: 1.0072272+rdf-ntriples.txt
Original Record: 1.0072272 +original-record.json
Full Text
1.0072272.txt
Citation
1.0072272.ris

Full Text

     MIXTURE FORMATION IN A PARTIALLY STRATIFIED CHARGE DIRECTLY INJECTED NATURAL GAS ENGINE  by  MALCOLM SHIELD  Master of Engineering (Hons.), Imperial College, University of London, 2003    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES  (Mechanical Engineering)          THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)    October 2011   © Malcolm Shield, 2011      ii  ABSTRACT A rapid compression machine was redesigned to allow the use of acetone UV laser diagnostics to investigate the mixture distribution that results from the injection of a methane partially stratified charge (PSC) and direct injection (DI) jet. A central composite test matrix was used to investigate the effect of relative injection timing and bulk charge air-fuel ratio upon the mixture distribution. Comparison was made between the distribution that resulted from a capillary injected PSC charge and a bespoke ‘sparkplug insert’ injected charge.  The capillary injected PSC jet was found to preserve a jet-like structure despite its interaction with the direct injection jet, while the effect of the DI jet upon the insert injected fuel was to encourage coalescence of the jets to form a largely homogeneous mixture at the point of injection and near the leading edge of the DI jet. The DI jet, with a weak bulk charge, served to reduce the fluctuations in relative air-fuel ratio compared to PSC injection into air; while the insert injected PSC charge exhibited increased fluctuation levels with advanced relative injection timing. The improved ignition of a partially stratified charge from the introduction of a weak bulk charge had been presumed to work through reductions in fuel concentration gradients, however the findings of the this work suggest that this works in unison with a decrease in fuel concentration fluctuations that increases ignition efficacy. The insert injected PSC fuel demonstrates scalar dissipation rates that are potentially too low to provide robust enough combustion for a viable partially stratified charge approach. The PSC insert engenders more mixing than a capillary injected PSC, but penetrates the DI jet less well. In all cases, and throughout the region of the interaction, there exists a finite probability of encountering pure fuel or the bulk fuel concentration that suggests mixing driven by engulfment rather than entrainment. The PSC ‘sparkplug’ insert offers better opportunity for mixing than the capillary injection and using a stochastic design approach should be pursued further to improve the performance of partially stratified charge combustion for natural gas engines.     iii  TABLE OF CONTENTS ABSTRACT .................................................................................................................................................................. ii TABLE OF CONTENTS................................................................................................................................................. iii LIST OF TABLES......................................................................................................................................................... vi LIST OF FIGURES....................................................................................................................................................... vii LIST OF SYMBOLS & ABBREVAIATIONS.................................................................................................................. xiii ACKNNOWLEDGEMENTS.......................................................................................................................................... xvi DEDICATION .......................................................................................................................................................... xviii CHAPTER 1 INTRODUCTION ............................................................................................................................ 1 1▪1 INTRODUCTION................................................................................................................................................. 1 1▪2 THE NEED ........................................................................................................................................................ 2 1▪3 THE DIRECTLY INJECTED PARTIALLY STRATIFIED CHARGE NATURAL GAS ENGINE ...................................... 4 1▪4 AIMS................................................................................................................................................................. 5 1▪5 OBJECTIVES...................................................................................................................................................... 5 1▪6 DISSERTATION STRUCTURE.............................................................................................................................. 6 CHAPTER 2 GASEOUS JETS, GAS ENGINES & POLLUTANT EMISSIONS ............................................ 7 2▪1 INTRODUCTION................................................................................................................................................. 7 2▪2 TURBULENT GASEOUS JETS ............................................................................................................................. 8 2▪3 NATURAL GAS ENGINES ................................................................................................................................ 15 2▪4 POLLUTANTS AND ENGINE EMISSIONS .......................................................................................................... 23 2▪5 OXIDES OF NITROGEN.................................................................................................................................... 24 2▪6 CARBON MONOXIDE ...................................................................................................................................... 26 2▪7 HYDROCARBON EMISSIONS............................................................................................................................ 27 2▪8 PARTICULATE MATTER .................................................................................................................................. 29 2▪9 HYDROCARBON COMBUSTION & CARBON DIOXIDE PRODUCTION ................................................................ 30 2▪10 PHOTOCHEMICAL SMOG & TROPOSPHERIC OZONE...................................................................................... 31 2▪11 HEALTH IMPLICATIONS OF ENGINE EMISSIONS ........................................................................................... 31 2▪12 CLIMATIC IMPLICATIONS OF ENGINE EMISSIONS ......................................................................................... 32 2▪13 CONCLUSIONS .............................................................................................................................................. 32 CHAPTER 3 LASER IMAGING TECHNIQUES & PLIF FUNDAMENTALS ............................................ 33 3▪1 INTRODUCTION............................................................................................................................................... 33 3▪2 BACKGROUND ................................................................................................................................................ 34 3▪3 COHERENT AND INCOHERENT PROCESSES..................................................................................................... 34 3▪4 VELOCITY TECHNIQUES ................................................................................................................................. 35 3▪5 RAYLEIGH AND MIE SCATTERING ................................................................................................................. 35 3▪6 RAMAN SCATTERING ..................................................................................................................................... 36    iv  3▪7 COHERENT ANTI-RAMAN SCATTERING.......................................................................................................... 38 3▪8 LASER INDUCED FLUORESCENCE ................................................................................................................... 39 3▪9 APPLICATIONS OF PLANAR LASER INDUCED FLUORESCENCE ....................................................................... 40 3▪10 TRACER CHOICES ......................................................................................................................................... 42 3▪11 ACETONE AS A FLUORESCENT MEDIUM....................................................................................................... 49 3▪12 CONCLUSIONS .............................................................................................................................................. 57 CHAPTER 4 EXPERIMENTAL SETUP & PROCEDURES........................................................................... 59 4▪1 INTRODUCTION............................................................................................................................................... 59 4▪2 THE RAPID COMPRESSION MACHINE ............................................................................................................. 60 4▪3 NEW RCM CYLINDER .................................................................................................................................... 61 4▪4 RCM FUELLING SYSTEM ............................................................................................................................... 61 4▪5 FLOW SEEDING............................................................................................................................................... 64 4▪6 LASER & OPTICAL SYSTEM............................................................................................................................ 65 4▪7 EVENT & TIMING CONTROL........................................................................................................................... 66 4▪8 TEST MATRIX................................................................................................................................................. 68 CHAPTER 5 REPEATABILITY & CALIBRATION TESTS.......................................................................... 79 5▪1 INTRODUCTION............................................................................................................................................... 79 5▪2 CAMERA TRIGGER REPEATABILITY ............................................................................................................... 80 5▪3 ICCD CALIBRATION....................................................................................................................................... 81 5▪4 FLUORESCENT SIGNAL CHECKS ..................................................................................................................... 84 5▪5 PRESSURE TRANSDUCER CALIBRATION ......................................................................................................... 87 5▪6 POLYTROPIC EXPONENT CALCULATION......................................................................................................... 88 5▪7 LASER SHEET PROFILE CORRECTION ............................................................................................................. 89 5▪8 OPTICAL DISTORTION CORRECTION............................................................................................................... 90 5▪9 PIXEL INTENSITY AIR-FUEL-RATIO CALIBRATION ........................................................................................ 91 5▪10 IMAGE BINNING ........................................................................................................................................... 92 5▪11 SEEDER CHARACTERISATION ....................................................................................................................... 94 5▪12 FLOW ENVIRONMENT IN RCM CYLINDER ................................................................................................... 95 5▪13 J43M INJECTOR CALIBRATION ..................................................................................................................... 95 5▪14 CONCLUSION ................................................................................................................................................ 96 CHAPTER 6 IMAGE CALIBRATION............................................................................................................... 97 6▪1 INTRODUCTION............................................................................................................................................... 97 6▪2 IMAGE PROCESSING PROCEDURES AND PRESENTATION................................................................................. 98 6▪3 CORRECTION PROCEDURES ............................................................................................................................ 99 6▪4 RESULTS....................................................................................................................................................... 100 6▪5 DISCUSSION.................................................................................................................................................. 100 6▪6 CONCLUSIONS .............................................................................................................................................. 111    v  CHAPTER 7 PARTIALLY STRATIFIED CHARGE, DIRECT INJECTION & INSERT FUEL JETS.. 112 7▪1 INTRODUCTION............................................................................................................................................. 112 7▪2 DATA PRESENTATION................................................................................................................................... 113 7▪3 INEFFECTUAL BACKGROUND CORRECTION.................................................................................................. 113 7▪4 PARTIAL STRATIFICATION WITH DIRECT INJECTION- FULL BORE IMAGING................................................ 116 7▪5 PARTIAL STRATIFICATION WITH CAPILLARY TUBE INJECTION .................................................................... 120 7▪6 PARTIAL STRATIFICATION WITH SPARK PLUG INSERT INJECTION................................................................ 127 7▪7 CONCLUSION ................................................................................................................................................ 130 CHAPTER 8 THE INTERACTION OF PARTIALLY STRATIFIED & DIRECT INJECTION FUEL JETS………….. ........................................................................................................................................................ 131 8▪1 INTRODUCTION............................................................................................................................................. 131 8▪2 BACKGROUND .............................................................................................................................................. 132 8▪3 MIXTURE DISTRIBUTION .............................................................................................................................. 133 8▪4 CAPILLARY INJECTED PSC IGNITION & INFLAMMATION POTENTIAL .......................................................... 137 8▪5 PSC INSERT INJECTION IGNITION & INFLAMMATION POTENTIAL ................................................................ 151 8▪6 CONCLUSIONS .............................................................................................................................................. 167 CHAPTER 9 ERROR AND UNCERTAINTY ANALYSIS ............................................................................ 169 9▪1 INTRODUCTION............................................................................................................................................. 169 9▪2 DISCOUNTING BUOYANCY ........................................................................................................................... 170 9▪3 CYCLIC INVARIANCE.................................................................................................................................... 170 9▪4 EXPERIMENTAL CONVERGENCE................................................................................................................... 171 9▪5 UNCERTAINTY ANALYSIS............................................................................................................................. 176 9▪6 CONCLUSIONS .............................................................................................................................................. 179 CHAPTER 10 CONCLUSIONS & RECOMMENDATIONS......................................................................... 180 10▪1 INTRODUCTION........................................................................................................................................... 180 10▪2 CONCLUSIONS ............................................................................................................................................ 181 10▪3 RECOMMENDATIONS & FUTURE WORK ..................................................................................................... 185 REFERENCES……………………………………………………………………………………………………..187 APPENDIX A RCM APPARATUS & RE-DESIGN ........................................................................................ 215 APPENDIX B RCM TIMING CONTROL SYSTEM...................................................................................... 240 APPENDIX C RELATIVE AIR-FUEL-RATIO PDFS.................................................................................... 261 APPENDIX D STATISTICAL TREATMENTS .............................................................................................. 271      vi  LIST OF TABLES TABLE 1-1 CORE CURRENT AND PROPOSED EMISSIONS STANDARDS FOR ‘LIGHT DUTY VEHICLE’ ROAD TRANSPORT ............................................................................................................................................................................. ..4 TABLE 2-1 CENTRE-LINE ROOT MEAN SQUARE TURBULENT FLUCTUATION VALUES FOR JET CONCENTRATION, AXIAL VELOCITY AND RADIAL VELOCITY...................................................................................................................... 13 TABLE 2-2 MAXIMUM ROOT MEAN SQUARE TURBULENT FLUCTUATION VALUES FOR CONCENTRATION .................. 14 TABLE 3-1 ADVANTAGES AND DISADVANTAGES TO COHERENT AND INCOHERENT LASER DIAGNOSTICS TECHNIQUES ............................................................................................................................................................................. 35 TABLE 3-2 PHYSICAL & THERMODYNAMIC PROPERTIES OF METHANE AND ACETONE GAS, ....................................... 50 TABLE 3-3 ACETONE EXCITATION SOURCES, ............................................................................................................... 50 TABLE 4-1 RCM TIMED EVENTS.................................................................................................................................... 66 TABLE 4-2 DATA POINT LOCATIONS IN DESIGN (X) AND EXPERIMENTAL SPACE (Ξ) ................................................... 70 TABLE 4-3 CONTROL BOX CAPILLARY INJECTED PSC EVENT TIMINGS....................................................................... 71 TABLE 4-4 CONTROL BOX INSERT INJECTED PSC EVENT TIMINGS.............................................................................. 71 TABLE 4-5 DATA POINT EXPERIMENTAL SETUP VALUES ............................................................................................. 71 TABLE 4-6 HOMOGENEOUS CHARGE PARTIAL PRESSURE (GAUGE) VALUES ASSUMING PRIOR AIR FLUSH. ............... 72 TABLE 4-7 MISCELLANEOUS EQUIPMENT SETTINGS .................................................................................................... 72 TABLE 4-8 CAPILLARY INJECTED PSC WITH DI EXPERIMENTAL TEST ORDER ............................................................ 75 TABLE 4-9 INSERT INJECTED PSC WITH DI EXPERIMENTAL TEST ORDER ................................................................... 78 TABLE 5-1 MICROGC NATURAL GAS COMPOSITION RESULTS .................................................................................... 85 TABLE 5-2 LEAST SQUARES FIT COEFFICIENTS FOR LASER SHEET PROFILE ................................................................ 90 TABLE 5-3 MIXTURE PROPERTIES BASED ON DATA IN TABLE 5-1 AND ORIGINAL GC DATA....................................... 92 TABLE 5-4 DICAM PRO ICCD NOISE PARAMETERS ..................................................................................................... 93 TABLE 6-1 ANTOINE EQUATION COEFFICIENTS FOR ACETONE,  REPRODUCED FROM [383] ....................................... 104 TABLE 8-1 CAPILLARY INJECTED PSC SAMPLE STATION LOCATIONS........................................................................ 141 TABLE 8-2 PROBABILITY OF MIXTURE FALLING WITHIN FLAMMABILITY LIMITS FOR CAPILLARY INJECTED PSC JET WITH DI AT STATIONS ONE TO FOUR ................................................................................................................. 145 TABLE 8-3 INSERT INJECTED PSC SAMPLE STATION LOCATIONS .............................................................................. 154 TABLE 8-4 PROBABILITY OF MIXTURE FALLING WITHIN FLAMMABILITY LIMITS FOR INSERT INJECTED PSC JET WITH DI AT STATIONS ONE TO FIVE............................................................................................................................ 160 TABLE 8-5 INCREASE IN PROBABILITY OF MIXTURE FALLING WITHIN FLAMMABILITY LIMITS FOR INSERT INJECTED PSC JET OVER CAPILLARY INJECTED JET AT STATIONS ONE AND FIVE ............................................................ 161 TABLE 9-1 PEARSON PRODUCT-MOMENT FOR CAPILLARY INJECTED PSC DI IMAGES ............................................. 170 TABLE 9-2 MAXIMUM ENTROPY FOR NON-ZERO MEAN NORMAL DISTRIBUTION, AFTER [411] ................................ 175 TABLE A-1 HPFS 7890 (KRF GRADE) PROPERTIES ..................................................................................................... 218 TABLE A-2 RCM FUELLING OPTIONS ........................................................................................................................ 229 TABLE A-3 RCM FUELLING VALVE SETTINGS........................................................................................................... 230 TABLE A-4 COMPONENT LISTING FOR RCM SPARK CIRCUIT .................................................................................... 239 TABLE A-5 INSTRUMENT LIST FOR RCM ................................................................................................................... 239     vii   LIST OF FIGURES  FIGURE 1-1 WORLD ENERGY CONSUMPTION BY SOURCE, ............................................................................................. 2 FIGURE 1-2 ENERGY GROWTH DUE TO INCREASED TRANSPORT DEMAND BY REGION 2006-2030,............................... 2 FIGURE 1-3 PARTIALLY STRATIFIED CHARGE WITH DIRECT INJECTION. ........................................................................ 4 FIGURE 2-1 THE VORTEX BALL MODEL OF A TRANSIENT PLUME BY TURNER,............................................................ 10 FIGURE 2-2 NORMALISED RADIAL R.M.S. PROFILES OF AXIAL VELOCITY (W), RADIAL VELOCITY (U) AND CONCENTRATION (C),........................................................................................................................................... 12 FIGURE 2-3 POWER REDUCTION FROM INCREASED BURN DURATION AND NECESSITATED SPARK ADVANCE, ........... 16 FIGURE 2-4 GHG REDUCTION POTENTIAL OF CNG ENGINES AGAINST GASOLINE EQUIVALENTS, ............................. 17 FIGURE 2-5 OPERATING ENVELOPE FOR AN SI GAS ENGINE, ....................................................................................... 19 FIGURE 2-6 LEAN LIMIT EXTENSION & BSFC REDUCTION (LEFT) AND NOX REDUCTION POTENTIAL (RIGHT) OF PSC ............................................................................................................................................................................. 21 FIGURE 2-7 FUEL EFFICIENCY OF LOAD CONTROL STRATEGIES,.................................................................................. 22 FIGURE 2-8 ENGINE POLLUTANT CONCENTRATIONS ACROSS THE RANGE OF OPERATING FAR’S,.............................. 24 FIGURE 2-9 FUEL JET STRUCTURE IN A COMPRESSION IGNITION ENGINE, ................................................................... 28 FIGURE 2-10 THE EFFECT OF OVERFUELLING ON EXHAUST HC CONCENTRATIONS, ................................................... 28 FIGURE 2-11 HIERARCHICAL NATURE OF HC COMBUSTION, ........................................................................................ 30 FIGURE 3-1 RO-VIBRONIC STRUCTURE OF THE RAMAN SPECTRUM,............................................................................. 36 FIGURE 3-2 CARS ADAPTED FROM [146] ....................................................................................................................... 39 FIGURE 3-3 MOLECULAR ACETONE, (CH3)2CO............................................................................................................ 49 FIGURE 3-4 DEACTIVATION PATHWAYS FOR THE FIRST EXCITED SINGLET, S1, OF ACETONE, ..................................... 49 FIGURE 3-5 ABSORPTION SPECTRUM OF ACETONE AT ROOM TEMPERATURE AND 1 ATM, ........................................... 51 FIGURE 3-6 ACETONE DISSOCIATION THROUGH Α-CLEAVAGE,.................................................................................... 52 FIGURE 3-7 JABLONSKI DIAGRAM FOR A ELECTRONICALLY EXCITED ORGANIC MOLECULE,...................................... 54 FIGURE 3-8 FLUORESCENT (SOLID LINE) AND PHOSPHORESCENT (BROKEN LINE) EMISSION SPECTRA FOR PURE ACETONE, ............................................................................................................................................................ 55 FIGURE 3-9 FLUORESCENT LIFETIME VARIATION OF ACETONE WITH PRESSURE, ........................................................ 57 FIGURE 4-1 SCHEMATIC OF RAPID COMPRESSION MACHINE (RCM) ............................................................................. 60 FIGURE 4-2 EXPLODED VIEW OF NEW RCM CYLINDER................................................................................................. 61 FIGURE 4-3 SCHEMATIC OF RCM FUELLING SYSTEM .................................................................................................... 62 FIGURE 4-4 PSC INSERT ............................................................................................................................................... 64 FIGURE 4-5 SECTION VIEW OF RCM SEEDER ................................................................................................................. 65 FIGURE 4-6 LASER SHEET FORMATION......................................................................................................................... 66 FIGURE 4-7 RCM EVENT TIMING DIAGRAM................................................................................................................. 67 FIGURE 4-8 RCM INSTRUMENTATION SCHEMATIC ........................................................................................................ 68 FIGURE 4-9 PARTIALLY STRATIFIED CHARGE JET AND DIRECT INJECTION JET INTERACTION ..................................... 69 FIGURE 4-10 CENTRAL COMPOSITE DESIGN FOR TEST MATRIX................................................................................... 70     viii  FIGURE 5-1 MEAN PISTON POSITION AND STANDARD DEVIATION OF POSITION FOR 4300, 4800 AND 5800 CAMERA TRIGGER COUNT TIMING ..................................................................................................................................... 80 FIGURE 5-2  ICCD FLAT FIELD IMAGE WITH MEAN INTENSITY AT 26 BITS, WITH (RIGHT) AND WITHOUT (LEFT) RANGE CROPPING............................................................................................................................................................. 81 FIGURE 5-3  MEAN AND STANDARD DEVIATION FOR THE ICCD DARK FIELD.............................................................. 83 FIGURE 5-4  CENTRAL ROW (LEFT) AND COLUMN (RIGHT) FOR ICCD FLAT FIELD SUBREGION .................................. 83 FIGURE 5-5  MEAN TEMPORAL-SPATIAL CORRELATION PLOTS FOR ICCD FLAT FIELD ................................................ 84 FIGURE 5-6  AIR NORMALISED IMAGES OF FLUORESCENT INTENSITY FROM NATURAL GAS (LEFT) AND NATURAL GAS WITH 1% ACETONE V/V (RIGHT). .......................................................................................................................... 86 FIGURE 5-7  UNSEEDED DIRECT INJECTION EVENT NORMALISED AGAINST QUIESCENT AIR CHARGE ........................ 87 FIGURE 5-8  POLYTROPIC EXPONENT PLOTS FOR AIR (LEFT) AND STOICHIOMETRIC CH4 MIXTURE (RIGHT)............... 88 FIGURE 5-9  LASER SHEET PROFILE; RECORDED (•) AND WITH A 4TH ORDER FIT (-) ..................................................... 89 FIGURE 5-10 CYLINDER OPTICAL DISTORTION............................................................................................................. 90 FIGURE 5-11  NORMALISED FUEL NUMBER COUNT VS SIGNAL INTENSITY................................................................... 92 FIGURE 5-12 SIGNAL-TO-NOISE RATIO FOR 2×2 AND 4×4 ON-CHIP BINNING AT DIFFERENT SIGNAL INTENSITIES & MAXIMUM SNR CURVE....................................................................................................................................... 93 FIGURE 5-13  ACETONE SEED CONCENTRATIONS......................................................................................................... 95 FIGURE 5-14  FLAME PROPAGATION FOR A STOICHIOMETRIC MIXTURE IN RCM CYLINDER (TIME LISTED AFTER SPARK) ................................................................................................................................................................. 95 FIGURE 5-15  J43M INJECTOR CHARACTERISATION DATA AND MODEL PREDICTION .................................................. 96 FIGURE 6-1 IMAGE POST-PROCESSING ROAD MAP....................................................................................................... 98 FIGURE 6-2 INTENSITY IMAGE FOR TEST MATRIX CENTRAL LOCATION .................................................................... 100 FIGURE 6-3 ACETONE FLUORESCENT YIELD VARIATION WITH TEMPERATURE AT 248NM EXCITATION, ................... 103 FIGURE 6-4  FLUORESCENT SIGNAL RATIO TEMPERATURE VARIATION FOR 308/248NM AND 308/266NM PAIRS. REPRODUCED FROM [258].................................................................................................................................. 103 FIGURE 6-5 HIGH MAGNIFICATION SHADOWGRAPH IMAGES OF JET: (A) NATURAL-GAS JET, FAR-FIELD (B) ACETONE SEEDED NG JET, FAR-FIELD; (C) NG JET, NEAR-FIELD; (D) ACETONE SEEDED NG JET, NEAR-FIELD .................. 105 FIGURE 6-6 CFD GAS INJECTION PREDICTIONS OF (A) TEMPERATURE, (B) DENSITY AND (C) VELOCITY .................. 107 FIGURE 6-7 THE TURBULENT ENERGY CASCADE ....................................................................................................... 109 FIGURE 7-1 DATA VISUALISATION SCHEMATIC.......................................................................................................... 113 FIGURE 7-2 BACKGROUND CORRECTION IMAGE ........................................................................................................ 114 FIGURE 7-3 RAW LIF IMAGE ...................................................................................................................................... 114 FIGURE 7-4 INEFFECTUAL BACKGROUND CORRECTED IMAGE ................................................................................... 115 FIGURE 7-5 ROLLING BALL EROSION, ........................................................................................................................ 115 FIGURE 7-6 4×4 AND 2×2 ON-CHIP BINNING CALIBRATION LINES, INSET DETAIL FOR LOW SIGNAL INTENSITY ...... 117 FIGURE 7-7 NORMALIZED R.M.S FLUCTUATIONS IN RAFR FOR DIRECT INJECTION WITH PSC AT RELATIVE TIME INDICATED ......................................................................................................................................................... 118 FIGURE 7-8 RAFR FOR PSC JET FROM CAPILLARY TUBE AT TIMES INDICATED ........................................................... 121     ix  FIGURE 7-9 PSC CAPILLARY INJECTED JET CENTRELINE COUNT LEVELS (BROKEN LINES INDICATE 95% CONFIDENCE)..................................................................................................................................................... 122 FIGURE 7-10 CAPILLARY INJECTED PSC JET PENETRATION; (•) PSC DATA (-) LEAST SQUARES FIT ........................ 123 FIGURE 7-11 NORMALIZED R.M.S FLUCTUATIONS IN RAFR FOR CAPILLARY INJECTED PSC AT TIMES INDICATED ..... 125 FIGURE 7-12 SPATIAL FUEL GRADIENTS FOR CAPILLARY INJECTED PSC AT TIMES INDICATED ................................. 126 FIGURE 7-13 JET DEVELOPMENT FOR INSERT INJECTED PSC AT TIMES INDICATED .................................................... 127 FIGURE 7-14 FUEL CONCENTRATION GRADIENTS FOR INSERT INJECTED PSC WITH BULK CHARGE RAFRS INDICATED ........................................................................................................................................................................... 129 FIGURE 7-15 BACKGROUND CHARGE EFFECT UPON INSERT JET MIXTURE DISTRIBUTION ........................................ 129 FIGURE 8-1 CENTRAL COMPOSITE TEST MATRIX ....................................................................................................... 132 FIGURE 8-2 RAFR DISTRIBUTION WITH CAPILLARY INJECTED PSC WITH DIRECTION INJECTION ............................. 134 FIGURE 8-3 CAPILLARY INJECTED PSC JET CENTRELINE IMAGE INTENSITY.............................................................. 135 FIGURE 8-4 RAFR DISTRIBUTION WITH INSERT INJECTED PSC WITH DIRECTION INJECTION .................................... 136 FIGURE 8-5 DEFLECTION OF CAPILLARY INJECT PSC JET .......................................................................................... 137 FIGURE 8-6 CAPILLARY INJECT PSC & DI JET RAFR GRADIENTS............................................................................. 139 FIGURE 8-7 CAPILLARY INJECTED PSC AND DI JET NORMALISED RAFR FLUCTUATIONS ........................................ 140 FIGURE 8-8 SAMPLE STATION LOCATIONS FOR CAPILLARY INJECTED PSC & DI JETS .............................................. 141 FIGURE 8-9 PROBABILITY DENSITY FUNCTIONS FOR CENTRAL CONDITION OF THE TEST MATRIX AT (A) STATION ONE, (B) STATION TWO, (C) STATION THREE AND (D) STATION FOUR. ...................................................................... 142 FIGURE 8-10 PROBABILITY DENSITY FUNCTIONS FOR CAPILLARY INJECTED PSC JET WITH DI AT STATION ONE FOR (A) RAFR=1.573 T=-3.53CAD AND (B) RAFR=1.5 AND T=0CAD....................................................................... 143 FIGURE 8-11 MAXIMA, MINIMA AND MEAN RAFR FOR ALL CAPILLARY INJECTED PSC TEST CONDITIONS AT (A) STATION ONE, (B) STATION TWO, (C) STATION THREE AND (D) STATION FOUR. .............................................. 144 FIGURE 8-12 CAPILLARY INJECTED PSC WITH DI JET FUEL CONCENTRATIONS BY VOLUME. ................................... 146 FIGURE 8-13 PROBABILITY OF ENCOUNTERING A MIXTURE WITHIN THE FLAMMABILITY LIMITS OF METHANE FOR INSERT INJECTED PSC........................................................................................................................................ 147 FIGURE 8-14 SCALAR DISSIPATION RATE FOR CAPILLARY INJECTED PSC WITH DI JET ............................................ 150 FIGURE 8-15 INSERT INJECT PSC & DI JET RAFR GRADIENTS.................................................................................. 151 FIGURE 8-16 INSERT INJECTED PSC AND DI JET NORMALISED RAFR FLUCTUATIONS ............................................. 152 FIGURE 8-17 SAMPLE STATION LOCATIONS FOR INSERT INJECTED PSC DI EVENTS.................................................. 154 FIGURE 8-18 RAFR PROBABILITY DENSITY FUNCTIONS FOR T=0CAD AT λBULK=1.5 (LEFT), λBULK=1.75 (CENTRE) AND λBULK=2 (RIGHT) FOR (A) STATION ONE, (B) STATION TWO AND (C) STATION THREE. ....................................... 156 FIGURE 8-19 RAFR PROBABILITY DENSITY FUNCTIONS FOR T=0CAD AT λBULK=1.5 (LEFT), λBULK=1.75 (CENTRE) AND λBULK=2 (RIGHT) FOR (A) STATION FOUR AND (B) STATION FIVE. ....................................................................... 157 FIGURE 8-20 RAFR PROBABILITY DENSITY FUNCTIONS FOR T=-5CAD (CENTRE) AND T=-3.53CAD (LEFT & RIGHT) AT λBULK=1.57 (LEFT), λBULK=1.75 (CENTRE) AND λBULK=1.93 (RIGHT) FOR STATION FIVE. ....................................... 157 FIGURE 8-21 MAXIMA, MINIMA AND MEAN RAFR FOR ALL INSERT INJECTED PSC TEST CONDITIONS AT (A) STATION ONE, (B) STATION TWO AND (C) STATION THREE.............................................................................................. 158     x  FIGURE 8-22 MAXIMA, MINIMA AND MEAN RAFR FOR ALL INSERT INJECTED PSC TEST CONDITIONS AT (A) STATION FOUR AND (B) STATION FIVE ............................................................................................................................. 159 FIGURE 8-23 EFFECT OF RELATIVE INJECTION TIMING ON MEAN RAFR FOR STATION ONE (LEFT) AND STATION FIVE (RIGHT) .............................................................................................................................................................. 160 FIGURE 8-24 INSERT INJECTED PSC WITH DI JET FUEL CONCENTRATIONS BY VOLUME. .......................................... 163 FIGURE 8-25 PROBABILITY OF ENCOUNTERING A MIXTURE WITHIN THE FLAMMABILITY LIMITS OF METHANE FOR INSERT INJECTED PSC........................................................................................................................................ 164 FIGURE 8-26 SCALAR DISSIPATION RATE FOR CAPILLARY INJECTED PSC WITH DI JET ............................................ 166 FIGURE 9-1 QUINTILE-QUINTILE PLOTS FOR MEAN PIXEL INTENSITY AT STATIONS 1 (LEFT) & 3 (RIGHT) FOR CAPILLARY INJECTED PSC-DI EVENT WITH BULK CHARGE RAFR  OF 1.5 (A), 1.75 (B) AND 2.0 (C) ............... 172 FIGURE 9-2 QUINTILE-QUINTILE PLOTS FOR MEAN PIXEL INTENSITY AT STATIONS 1 (LEFT) & 3 (RIGHT) FOR INSERT INJECTED PSC-DI EVENT WITH BULK CHARGE RAFR OF 1.5 (A), 1.75 (B) AND 2.0 (C).................................... 173 FIGURE 9-3 NORMALISED ENTROPY IMAGES OF LIF INTENSITY FOR CAPILLARY INJECTED PSC WITH DI AT T=0CAD AND (A) RAFR = 1.5 AND 51 REPEATS, AND (B) RAFR = 1.75 AND 153 REPEATS BASED ON A NORMAL DISTRIBUTION. ................................................................................................................................................... 176 FIGURE 9-4 NORMALISED ENTROPY IMAGES OF LIF INTENSITY FOR CAPILLARY INJECTED PSC WITH DI AT T=0CAD AND (A) RAFR = 1.5 AND 51 REPEATS, AND (B) RAFR = 1.75 AND 153 REPEATS BOTH BASED ON POISSON STATISTICS......................................................................................................................................................... 176 FIGURE 9-5 UNCERTAINTY IN RELATIVE AIR-FUEL RATIO FOR INSERT INJECTED PSC WITH DI AT T=0CAD AND (A) λBULK=1.5 (B) λBULK=1.75 AND (C) λBULK=2.0......................................................................................................... 178 FIGURE A-1 RCM SCHEMATIC ................................................................................................................................... 216 FIGURE A-2 OLD ACRYLIC RCM CYLINDER .............................................................................................................. 217 FIGURE A-3 RCM QUARTZ CYLINDER ....................................................................................................................... 218 FIGURE A-4 RCM QUARTZ CYLINDER HEAD (WINDOW)........................................................................................... 219 FIGURE A-5 PROBLEMATIC ELEMENTS IN QUARTZ CYLINDER FEA .......................................................................... 220 FIGURE A-6 ELEMENT ASPECT RATIOS IN AREA OF CONCERN FOR QUARTZ CYLINDER FEA ................................... 220 FIGURE A-7 ELEMENTAL JACOBIAN VALUES FOR QUARTZ CYLINDER FEA .............................................................. 221 FIGURE A-8  RCM QUARTZ CYLINDER FOS VALUES BASED ON A MOHR-COULOMB YIELD CRITERIA FOR COMPRESSION RATIO 5.33(A), 5.19(B), 8.18(C) & 13.78(D) .............................................................................. 223 FIGURE A-9 RCM PORT PLATE .................................................................................................................................. 224 FIGURE A-10 RCM ELECTRODE MOUNT.................................................................................................................... 225 FIGURE A-11 RCM PISTON......................................................................................................................................... 226 FIGURE A-12 RCM CYLINDER EXPLODED VIEW........................................................................................................ 227 FIGURE A-13 RCM COMPRESSION PLATE DRAWING ................................................................................................. 227 FIGURE A-14 RCM FUELLING SYSTEM ...................................................................................................................... 228 FIGURE A-15 RCM FUELLING PANEL VALVING, PIPING & VALVE NUMBERING ....................................................... 229 FIGURE A-16 RCM DI TIP DRAWING ......................................................................................................................... 231 FIGURE A-17 RCM PSC INJECTOR MOUNT................................................................................................................ 232 FIGURE A-18 ACETONE SEEDER SECTION VIEW......................................................................................................... 233     xi  FIGURE A-19 ACETONE SEEDER: BODY...................................................................................................................... 234 FIGURE A-20 ACETONE SEEDER: BOTTOM FLANGE ................................................................................................... 234 FIGURE A-21 ACETONE SEEDER: TOP FLANGE........................................................................................................... 235 FIGURE A-22 ACETONE SEEDER: CUP ........................................................................................................................ 235 FIGURE A-23 ACETONE SEEDER: GUIDE..................................................................................................................... 236 FIGURE A-24 ACETONE SEEDER: PIPE ASSEMBLIES ................................................................................................... 236 FIGURE A-25 ACETONE SEEDER: TOP WELD-UP........................................................................................................ 237 FIGURE A-26 ACETONE SEEDER: ASSEMBLY DRAWING............................................................................................. 237 FIGURE A-27 RCM SPARK CIRCUIT, CIRCUIT DIAGRAM ........................................................................................... 239 FIGURE B-1 INFORMATION FLOW IN RCM CONTROL................................................................................................. 241 FIGURE B-2 ENCODER SIGNAL GATE, CIRCUIT DIAGRAM.......................................................................................... 243 FIGURE B-3 PULSE TRAIN DISTRIBUTION, CIRCUIT DIAGRAM ................................................................................... 244 FIGURE B-4 RESET SIGNAL DISTRIBUTION, CIRCUIT DIAGRAM ................................................................................. 245 FIGURE B-5 ENCODER PULSE TRAIN RECEIVING, CIRCUIT DIAGRAM ........................................................................ 245 FIGURE B-6 PULSE COUNTER AND COMPARATOR, CIRCUIT DIAGRAM ...................................................................... 247 FIGURE B-7 LOGIC PULSE GENERATION, CIRCUIT DIAGRAM ..................................................................................... 247 FIGURE B-8 LOGIC SIGNAL GENERATION, CIRCUIT DIAGRAM ................................................................................... 248 FIGURE B-9 START COUNTER-GATE A, PULSE CIRCUIT............................................................................................. 249 FIGURE B-10 START COUNTER-GATE B, PULSE CIRCUIT ........................................................................................... 250 FIGURE B-11 LATCH & RESET CIRCUITS, PULSE CIRCUIT.......................................................................................... 250 FIGURE B-12 SMOOTH CAPACITOR CIRCUITS, PULSE CIRCUIT................................................................................... 251 FIGURE B-13 START COUNTER-GATE A, LOGIC CIRCUIT........................................................................................... 252 FIGURE B-14 START COUNTER-GATE B, LOGIC CIRCUIT ........................................................................................... 252 FIGURE B-15 LATCH CIRCUIT..................................................................................................................................... 253 FIGURE B-16 SIGNAL RECEIVING, LOGIC CIRCUIT ..................................................................................................... 253 FIGURE B-17 SMOOTH CAPACITOR CIRCUITS, LOGIC CIRCUIT................................................................................... 254 FIGURE B-18 COUNTER CIRCUIT, START, INJECTOR CIRCUIT..................................................................................... 255 FIGURE B-19 COUNTER CIRCUIT, DURATION, INJECTOR CIRCUIT .............................................................................. 255 FIGURE B-20 J-K FLIP FLOP INJECTOR CIRCUIT ......................................................................................................... 256 FIGURE B-21 RESET CIRCUIT, INJECTOR .................................................................................................................... 256 FIGURE B-22 SIGNAL RECEIVING, INJECTOR CIRCUIT ................................................................................................ 257 FIGURE B-23 SMOOTHING CAPACITORS, INJECTOR CIRCUIT ...................................................................................... 257 FIGURE B-24 SIGNAL DISTRIBUTION CIRCUITS .......................................................................................................... 260 FIGURE B-25 RCM CONTROL CIRCUIT ENCLOSURE .................................................................................................... 260 FIGURE C-1 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR CAPILLARY INJECTED PSC EVENT WITH DI AT STATION ONE (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-1) ..................................................................................................... 262     xii  FIGURE C-2 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR CAPILLARY INJECTED PSC EVENT WITH DI AT STATION TWO (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-1) ..................................................................................................... 263 FIGURE C-3 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR CAPILLARY INJECTED PSC EVENT WITH DI AT STATION THREE (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-1) ..................................................................................................... 264 FIGURE C-4 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR CAPILLARY INJECTED PSC EVENT WITH DI AT STATION FOUR (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-1) ..................................................................................................... 265 FIGURE C-5 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR INSERT INJECTED PSC EVENT WITH DI AT STATION ONE (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-3) ............................................................................................................. 266 FIGURE C-6 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR INSERT INJECTED PSC EVENT WITH DI AT STATION TWO (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-3) ............................................................................................................. 267 FIGURE C-7 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR INSERT INJECTED PSC EVENT WITH DI AT STATION THREE (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-3) ............................................................................................................. 268 FIGURE C-8 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR INSERT INJECTED PSC EVENT WITH DI AT STATION FOUR (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-3) ............................................................................................................. 269 FIGURE C-9 RELATIVE AIR FUEL RATIO PROBABILITY DENSITY FUNCTIONS FOR INSERT INJECTED PSC EVENT WITH DI AT  STATION FIVE (FOLLOWING STANDARD IMAGE PRESENTATION PROTOCOL USED IN MAIN TEXT, AND STATION LOCATION AS PER TABLE 8-3) ............................................................................................................. 270 FIGURE D-1 (A) CIRCUMSCRIBED CENTRAL COMPOSITE DESIGN (B) INSCRIBED CENTRAL COMPOSITE DESIGN, TEST DOMAIN SHOWN IN GREY, DATA POINTS AS BLACK DOTS ............................................................................... 273     xiii  LIST OF  SYMBOLS & ABBREVAIATIONS Nomenclature: Roman B Bayesian entropy P data 95% confidence interval B0 exit buoyancy Pa pa ambient pressure C concentration Pr pressure ratio c concentration volume flux  mean concentration Q normalising co-ordinate  exit concentration rc volume (compression) ratio  r.m.s. concentration R gas constant Pearson product moment  centre-line r.m.s. concentration r radius d diameter r0 exit radius D binary diffusion coefficient s solid angle d diameter S0 singlet ground state de equivalent diameter Sf fluorescent intensity dps pseudo-diameter S1 first excited singlet state f mixture fraction sr steridian fc centre-line mixture fraction T1 first excited triplet state g gravity u (radial) velocity g′ reduced gravity u′ radial velocity fluctuation  Plank’s constant ut turbulent velocity H entropy  mean radial velocity k kinetic energy  centre-line radial r.m.s. velocity kjc jet constant for concentration  r.m.s. radial velocity fluctuation kjw jet constant for velocity w axial velocity Ks mass flow rate constant w′ axial velocity fluctuation  mean kinetic energy squared  mean exit velocity L largest eddy lengthscale  mean centre-line velocity lI integral length-scale  r.m.s. axial velocity fluctuation lm momentum buoyancy ratio length scale x design space variable M0 exit momentum Y mass fraction  mass flow rate z axial co-ordinate  exit mass flow rate z* dimensionless axial co-ordinate Ni number of incident photons Zt jet penetration      xiv    Nomenclature: Greek α diffusivity µ dynamic viscosity αe entrainment velocity µ population mean αT turbulent diffusivity µeff effective viscosity Γ jet penetration constant ν kinematic viscosity Δθc combustion duration ξ experimental variable Δν frequency shift (in wavenumber) ρ0 exit density ε turbulent kinetic energy dissipation rate ρa ambient density η dimensionless radial co-ordinate,   σ        population standard deviation η Kolomogorov lengthscale τ timescale κ Kleinstein decay constant τI integral time-scale λ relative air fuel ratio  τk Kolomogorov time-scale λ profile width ratio Φ equivalence ratio λ wavelength χ scalar dissipation rate λ relative fuel-air-ratio                                                        xv    Abbreviations AC alternating current nmHC non-methyl hydrocarbon ATDC after top dead centre NG natural gas BMEP break mean effective pressure NOx oxides of nitrogen bsfc break specific fuel consumption OHD- RIKES optically heterodyned Raman induced Kerr effect spectroscopy BTDC before top dead centre PAH polyaromatic hydrocarbon CAD crank angle degrees PAN peroxyacetyl nitrate CARB California Air Resources Board PARS photoacoustic Raman spectroscopy CARS coherent anti-Raman scattering PFC perflurocarbon CCD charge coupled device PIV particle image velocimetry CI compression ignition PM particulate matter CI95% 95% confidence interval PMMA polymethylmethacrylate (acrylic) CNG compressed natural gas ppm parts per million CO carbon monoxide Pr Prandtl number CoV coefficient of variation PSC partially stratified charge CSRS Stokes-shifted Raman scattering PWM pulse width modulation DC direct current R&D research & development DI direct injection RC resistor-capacitor DISC direct injection stratified charge RCARS rotational-CARS EGR exhaust gas recirculation RCM rapid compression machine FAR fuel-air-ratio Re Reynolds number FWHM full width half maximum RELIEF Raman excitation plus laser induced electronic fluorescence GDI gasoline direct injection RFAR relative fuel air ratio HC hydrocarbon Ri Richardson number HCCI homogeneous charge compression ignition RIKES Raman-induced Kerr effect spectroscopy HFC hydrofluorcarbon RON research octane number HHV higher heating value RPV reaction progress variable HRR heat release rate Sc Schmidt number IMEP indicated mean effective pressure SI spark ignition IPCC Intergovernmental Panel on Climate Change SNR signal-to-noise ratio KrF krypton-fluorine (laser) SRGS stimulated Raman gain spectroscopy LDA laser Doppler anemometry SRLS stimulated Raman loss spectroscopy Le Lewis number SRS spontaneous Raman spectroscopy LIF laser induced fluorescence TTL transistor transistor logic M third body (chemical reactions) UHC unburnt hydrocarbon MBT mean timing for best torque UV ultra violet MW molecular weight VOC volatile organic compound Nd:YAG neodymium-doped yttrium aluminium garnet VVT variable valve timing   w.r.t. with respect to          xvi  ACKNNOWLEDGEMENTS It is rare that a journey as long as PhD follows the course originally set out or delivers the expected destination. My years at UBC are no exception. It is a misnomer to presume that Doctoral studies in engineering are the endeavour of an individual, in a single discipline, in a single technicality. My studies have taken me from mechanical design, to electronics, quantum physics and combustion. In doing so I have spoken to, met, worked with and relied on many people. Space does not allow me to thank them all, nor to be earnest enough about those whom I can mention.  My initial thanks must go to Dr Robert Evans for accepting me into his research group and in doing so allowing me to become part of UBC, an institution that introduced me to Vancouver- a city I now call home. I would like to thank Dr Martin Davy for having the vision to bring laser diagnostics to UBC, a field with huge scope for the individual researcher and for the wider research community. My gratitude also goes to Drs. Steven Rogak and Kendal Bushe for their acumen when called upon. I am also very much indebted to Glenn Golly for his patience in teaching a mechanical engineer about electronics; Sean Buxton for helping bring those teachings to life; Roland Genschorek for making sense of my engineering drawings; and Eric Wilson and Markus Fengler for their mechanical insight. My thanks must also go to those at Imperial College London who were responsible for laying the foundations of my engineering faculty. I would also like to offer my sincere thanks to Brian Beck for allowing me to pursue and complete my studies while continuing my professional development; and to Dr. Meryn Bowen for being, and continuing to be, a stalwart friend who is always happy to lend an ear.  No matter how dark the hour I have consistently been granted a little light from some choice fellow Graduate Students. Over the years each of Ed Chan, Dave Gorby, Dave Williams, James Saunders, Andrew Mezo, Jean Logan, and Erik Kastanis have provided humour and levity that will forever punctuate my time at UBC. On occasion they have even been known to provide some of astute insights (and sometimes less so).  At every turn and in every stage of my life I have been graced with the unqualified and unwavering support of my family. They have provided me with every opportunity and all the support I needed to succeed- a luxury to which I am not blind. My mother has, from the outset, instilled in me a set of principles which I drawn on daily and for which I am a better person; she has listen to the minutiae of my work with patience and understanding, despite not understanding what exactly I do. My step-mother has taught me to always be philosophical and look for happiness; and despite only a limited amount of time with my Father I learned from him the importance of application and all that it can bear.   The work presented herein is all my own with the exception of the CFD computations and results presented in Chapter 6, for which I offer Mr Ehsan Faghani my sincere thanks. The insight they provided helped further my understanding of gaseous injections.    xvii  DEDICATION             It is a mistake to suppose that men succeed through success; they much more often succeed through failures. Precept, study, advice and example could never have taught them so well as failure has done.  Samuel Smiles, 1859  1  Chapter 1  INTRODUCTION   1▪1 INTRODUCTION This opening chapter serves only to put the rest of the manuscript in context. The greater need for research into improved energy efficiency is presented, with an emphasis on the need for more efficient and cleaner road transport. A technology to facilitate exactly these is outlined- that of partially stratified charge combustion. An enhanced understanding of this technology and its potential integration into a direct injection engine are sought through the current laser diagnostics study, the aims of which are made explicit with a clear and measureable set of objectives. The chapter concludes with an outline of the dissertation structure. ▪ Chapter 1 ▪ Introduction ▪ 2  1▪2 THE NEED Energy is at the very core of society, particularly those societies which form the countries of the OECD1. The primary sources of energy open to us, as a race, are few: solar, nuclear and fossil derived; any other ‘sources’, including all ‘renewables’, can be tied back to these primary sources. Currently fossil fuels account for 80% of world energy demand (which totals 130-140GWh annually), cf. Figure 1-1, resulting in anthropogenic CO2 emissions of 29 Gt/yr [1].   Figure 1-1 World Energy Consumption by Source,  after [2].  Transport accounts for 23% of these emissions [1] and uses 61% of all produced oil, where gasoline avgas and diesel (including fuel bunker) account for 26%, 6% and 29% respectively [3]. Transport has been wedded to oil since the demise of coal; however, climate change and the scourge of urban pollution are forcing changes upon road transport. Rail, marine and air transport, although expected to improve, are not under the same pressures.  World transport demand is set to grow by 45% by 2030 [2] and will account for three-quarters of total oil consumption increase [2]. Only a small fraction of this growth is expected to come from the OECD, cf. Figure 1-2.     Figure 1-2 Energy Growth Due to Increased Transport Demand by Region 2006-2030,  after [2].                                                         1 Organisation for Economic Co-operation and Development Total 767Mto e ▪ Chapter 1 ▪ Introduction ▪ 3 Economic growth and improvement of the living standards are inexorably linked to increased energy demand, which until recently has been made with scant regard for the future, let alone regard for a sustainable future. Car consumers are their own fiduciaries and have forced the auto-industry to levels of efficiency and end-product performance which are the envy of most other manufacturers. Customer expectation will not support a move away from these standards in the name of pollution reduction, climate mitigation or sustainability. The International Energy Agency proposes in their BLUE Map scenario a cut in emissions of 50% (based on 2005 values) by 2050 to stabilise atmospheric CO2 at 450- 520ppm. These gains are to come from a range of the approaches, though it is the Second Generation transport solutions which must satisfy the most demanding consumers while addressing the emissions needs of years to come.  The technology to achieve these reductions is under development, some of which are nearing commercialization. Hybrid-electric cars are now widely available; and there is renewed vigour from Ford (Focus RV), Nissan (Leaf), Mitsubishi (MiEV), Subaru (G4e), Volvo (C30 electric) and Tesla (Roadster) in fully electric vehicles which Deutsche-Bank believe [3] to be the ‘game changing technology’ that will spell the end for oil and gas. Biofuels offer promise, although their exact benefit strongly depends on their nature: sugar-cane and corn-derived ethanol are not the same entity with respect to their carbon footprint. The touted hydrogen economy, the elixir of the sustainably growing economy, is still decades away from being realised and suffers a significant problem in that hydrogen is not an energy source but an energy carrier and arguably, some claim, an inefficient one at that. Accepting this, hydrogen still offers the potential to meet the toughest targets of BLUE Map. In moving towards these targets, natural gas (NG) offers significant potential. Oil has a current reserve to production ratio of around 40 years, while NG offers 60 [4] reaching 130 years [2] if flaring and venting are curtailed. Further, increased production capacity is likely to swell the capacity margin to about 27% (compared to ~12% presently) [2].  The larger technical and commercial problems of distribution and onboard storage need still to be overcome if the the potential benefits of a move to NG-powered transport ar to be realised. Natural gas technology currently offers CO2 reductions of ~20% compared to gasoline and provides a technological stepping stone towards hydrogen combustion.  Setting aside the climatic benefits, natural gas offers, in the short term, the most significant route to consistently meeting emissions regulations (which as of 2009 also regulate CO2 on a manufacturer fleet-wide basis in the USA). The targets to be met by the auto- manufacturers are outlined in Table 1-1, which summarizes an expansive and complex set of regulation for light duty vehicles (those up to about 3500 kg). Optimized natural gas engines can exploit the inherent cleanliness of gas to meet these standards with an ease arguably not available from the incremental steps being made on conventionally fuelled spark ignition (SI) and compression ignition (CI) engines.   ▪ Chapter 1 ▪ Introduction ▪ 4 NOx  (mg/km) CO  (mg/km) PM  (mg/km) nmHC  (mg/km) CH2O  (g/km) CO2  (g/km) Regulatory Standard Pertaining to Model Year Vehicle Type SI CI SI CI SI CI SI CI SI CI SI CI   CARB2 Low Emission Vehicle II [5]   2009 onwards <3864kg 31 --- 2100 --- 6.253 --- 47 --- 9.4 --- 2024 ---   Euro 5 [6]   2009-2014 <3500kg 82 280 2270 740 5 5 1085 --- --- --- --- --- Euro 6 [6]   2014 onwards   <1760kg 82 125 2270 740 5 5 1085 --- --- --- --- --- Table 1-1 Core Current and Proposed Emissions Standards for ‘Light Duty Vehicle’ Road Transport   Engine development trends are trying to bring together, into a common platform, the benefits of SI engines (low NOx and PM) and those of SI engines (low HC and CO). To meet emissions demands across the full operating envelope, future engine architecture is likely to be a combination of enhanced SI and CI technology, with further additions from technology like HCCI and partial stratification. 1▪3 THE DIRECTLY INJECTED PARTIALLY STRATIFIED CHARGE NATURAL GAS ENGINE Striving for emissions reductions and efficiency improvements, the research group of Prof. R L Evans has studied and developed a partially stratified charge, PSC, combustion system for natural gas SI-ICE’s [7]. The approach uses a modified spark plug to inject a small amount of fuel (about 5% w/w of the total charge) adjacent to the plug electrodes, so generating a mixture that is locally rich. The bulk cylinder charge then comprises an ultra-lean mixture that would not ignite if an unmodified sparkplug were used.   Figure 1-3 Partially Stratified Charge with Direct Injection.                                                         2 California Air Resources Board 3 Beyond 80,000 km durability. 4 Fleet-wide value, dropping to 133 mg/km by 2015. 5 Total allowable tHC’s 160 mg/km. ▪ Chapter 1 ▪ Introduction ▪ 5  The modified sparkplug ignites the rich pilot charge which burns vigorously enough to inflame the remaining bulk charge. Load control may be achieved through varying the bulk charge air-to-fuel ratio, although the current work is concerned with an approach capable of providing still better control- direct injection. Direct injection, of any fuel, will provide volumetric efficiency gains and offers the potential of throttless operation and the elimination of pumping losses. The work herein is presented as a study of the fundamental mixture distribution that results from combining partial stratification with direct injection. It is intended as a step towards explaining the phenomenological effects reported by D. Gorby [8] in his study of DI-PSC in a single-cylinder research engine. 1▪4 AIMS To Establish The Spatial Fuel Distribution Within The Cylinder of a Rapid Compression Machine Which Results From Both The Direct Injection & Partial Stratification of Natural Gas  The current study aims to understand the interaction of the partially stratified fuel jet with the jet generated by the direct injection of natural gas in a rapid compression machine. The study will achieve this through a detailed identification of the in-cylinder fuel distribution that results from the combined partial and direct-injection of fuel. The findings will be used to expand upon, and draw together, earlier work [8] on the utility of combined PSC-DI natural gas fuelling. 1▪5 OBJECTIVES The objectives to support the stated aim of this study are:  To quantitatively map the fuel air distribution which results from the partial- and direct-injection of natural gas into the cylinder of a rapid compression machine (RCM).  To identify the level and extent of the fluctuations in fuel concentration at the location of the spark electrodes in the RCM.  To understand the extent and role played by concentration gradients in the formation and ignition of the PSC fuel plume.  To understand the differences in the fuel distribution generated by a bespoke partially stratified charge (PSC) spark-plug ‘insert’ and a capillary tube injected PSC jet.  To understand the extent and implications of the inherent uncertainties associated with laser induced fluorescence imaging of acetone-doped natural gas jets in engine research. ▪ Chapter 1 ▪ Introduction ▪ 6 1▪6 DISSERTATION STRUCTURE Chapter 2 offers the reader background information on laminar and turbulent jets, pollutant formation in engines (with a discussion not limited to natural gas engines) and a basic understanding of natural gas engines. Laser diagnostics provide an expansive set of experimental techniques, each suited to providing slightly different information. Chapter 3 provides an overview of these techniques and their application, while a more substantial justification and explanation of laser induced fluorescence is made in the second half. Chapter 4 offers the reader the most salient details of the experimental apparatus used in the current study and outlines the experimental procedures used to obtain the results of Chapters 6-8. Chapter 5 outlines the wide range of calibration experiments undertaken to ensure that the results of the LIF study are of the highest quality possible of the experimental apparatus. Chapter 6 goes on to highlight some of the difficulties of establishing a quantitative LIF study for a highly underexpanded jet. Chapter 7 details the findings from a central composite test matrix that investigates the mixture distribution within the RCM cylinder for a range of relative direct injection injection timings and bulk charge air fuel ratios. The Chapter also establishes some basic behaviour of PSC injections made from a capillary tube and bespoke PSC ‘insert’. Chapter 8 repeats the same central composite test matrix to consider for a smaller region and in more detail the area where there is the interaction of the direct injection jet with that of the partially stratified charge produced from a capillary and contrasts it to that produced from the ‘insert’. Chapters 9 & 10 offer an assessment of the uncertainty in the experimental findings and some concluding remarks on the most important findings and recommended future work.  Appendices A-D present, respectively, the detailed redesign and design work undertaken to allow the RCM to be used for the current study, the timing and control hardware designed and implemented for the study, the relative air-fuel ratio probability density functions that resulted at a number of locations within the cylinder for the nine conditions of the test matrix and some fundamental information pertaining to the statistics used to assess the experimental results.  7 Chapter 2   GASEOUS JETS, GAS ENGINES & POLLUTANT EMISSIONS   2▪1 INTRODUCTION Chapter 2 aims to bring the reader through what, at first sight, may appear to be three disparate themes, but which are in fact closely related. The first portion of the chapter establishes the fundamental principles of gaseous jets and interaction with their surroundings. It is this interaction and mixing which forms the ignitable mixture within a natural gas engine, and it is the various features and designs of natural gas engines which are considered in the central portion of this chapter.  The chapter concludes with a discussion of the major pollutants to be expected from natural gas engines and the importance with respect to human health and climate change. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 8 2▪2 TURBULENT GASEOUS JETS The use of direct injection necessitates a fundamental understanding of turbulent free jets. The current discussion is limited to gaseous and submerged liquid jets. In general the flow from the nozzle of the injector used in the current work is fully turbulent (Re ~ 5×105), choked and under-expanded [9].  2▪2▪1 INCOMPRESSIBLE JET STRUCTURE & DEVELOPMENT Mass entrainment around the periphery of turbulent jets causes them to grow at, beyond the development region, a set half-angle. Early experimental work on this growth [10-13] confirmed the self-similarity first proposed for turbulent flows by Zel’Dovich [14], though there were still inconsistencies in the data. These centred on the failure of the turbulence intensities to attain self-similarity, and motivated the work of Wygnanski and Fiedler [15], whose hot-wire measurements provided the reference profiles for mean velocity, velocity fluctuation and turbulent stress for many years. Wygnanski & Fiedler’s work, as in previous research, made use of a virtual origin (at a distance up- or downstream of the nozzle exit) to circumvent the problems of specifying an initial condition, and allowed the simple hyperbolic form for the axial velocity of Eq. 2.1   Eq. 2.1 where de is the equivalent diameter first introduced by Thring & Newby [16] as d(ρ0/ρa) ½ accounts for density differences between the jet and its surroundings. However, concerns about such treatment of the source were first raised by Baker [17] when he showed Wygnanski & Fiedler’s formulation failed to preserve axial momentum. Using laser Doppler anemometry (LDA), Capp et al. [18] identified the source of the error as facility related, which motivated Schneider [19] to re-examine the theory of jets and show that the integrated momentum at any axial station is constant and equal to the rate at which it was added to the source. This extended understanding has led George [20] to revise the original implicit assumption of self-similarity, and assert that “the self-preserving state obtained [is] in fact … uniquely determined by the initial conditions.”. That is, jets cannot ‘forget’ their origins and will always hold a unique, though small, dependence upon their initial (source) conditions, primarily Reynolds number and velocity profile. An extensive re-examination of the experimental literature by Carazzo et al. [21] supports George’s work on local self-preservation, though it suggests that at large enough distances (where measurements have yet to be made) jets and plumes do indeed forget their initial condition and become fully self-similar with a non-Gaussian profile. Mi et al. [22] propose that before this state is achieved the source of local self- preservation is in the intermittent large-scale coherent structures of the flow. Close to the source such coherent structures have not formed, while at intermediate distances, large scale structures appear, but do ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 9 so only intermittently [23] and are still aware of their origin. Far enough away from the source these intermittent structures become permanent and self-similarity is obtained.  The evolution of velocity was shown by Wygnanski and Fiedler [15] to progress towards self-similarity in a stepwise fashion “since the energy is transferred from the mean motion directly to w' [axial] fluctuations, and only pressure-velocity-gradient correlations transfer the energy further to other components of the turbulent motion…”, i.e. self-similarity can be obtained only after a balance is achieved in the mean flow velocity and its fluctuation. The turbulence of the jet is heavily anisotropic as far as 100 diameters downstream, and this longitudinal preference skews the energy towards the lower end of the turbulent energy spectrum and highlights the preferential transfer of energy from the mean motion to  (mean axial velocity fluctuation) before  (mean radial velocity fluctuation). The nature of the anisotropy may reasonably be expected to depend upon the initial conditions and so ultimately affect the form of the energy redistribution. This is contradicted somewhat by Papanicalaou & List [24] who find the mean and fluctuating velocities to converge to self-similarity at the same point about 50 diameters downstream. Consideration of the evolution of the Reynolds stresses, reduced gravity, and kinetic energy by George [20] support the assertion of a stepped, and unequal, progress amongst the flow variables. The nature of the energy transfer affects the rate at which self-similarity is asymptotically approached, with the nature of the energy transfer equations shown in [20] to be a function of the initial conditions, i.e. source Reynolds number. The evolution of the flow may vary with changes in the profile shape and/or with the relative width of the buoyancy and velocity profiles, i.e. the local turbulent Prandtl number (the ratio of viscous diffusion to thermal diffusion). The evolution is, in general, seen to progress from a situation which is momentum governed (Prt~1) near the source to one in which vorticity transport dominates (Prt~0.5) [21]. Across the jet, however, the energy and Reynolds stress balances are affected differently. Near the core, balances are convection driven, while towards the jet boundary the balance is primarily the result of the production, dissipation and pressure-strain rate interactions in the Navier- Stokes equations [25]. The entrainment rate of a jet, which is in essence the result of all the interactions discussed above, is the primary process governing the mixing of fuel and oxidizer in a directly injected engine cylinder.   Considering the macroscopic details of a jet and taking a priori that the appropriate scaling is based on nozzle exit momentum [19, 26] allows an analysis of the problem that includes density gradients, compressibility and under-expansion [9]. Indeed, this a priori assumption is shown in [9] to be fully adequate for fuel jets, given that the opening transient is short compared to the duration of injection. The starting nature of a plume is most often modelled using the approach of Turner [27] where a quasi-steady- state jet is headed by a ‘vortex-ring’, cf. Figure 2-1. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 10  Figure 2-1 The Vortex Ball Model of a Transient Plume by Turner,  after [27].   Turner’s model is based upon buoyant atmospheric plumes, where an initial buoyant thermal rises and is subsequently ‘fed’ by the trailing quasi-steady-state plume that supplies about half of the mixed fluid in the ‘cap’ (the rest being entrained over the front of the cap). Assuming a spherical vortex and Gaussian velocity profile in the jet it can be shown that [27] the speed of advance for the cap is approximately 0.6 times that of the maximum velocity of the jet behind it, with a hyperbolic decrease in velocity with downstream location. It would be reasonable to assume that the dominant role of buoyancy in Turner’s model limits its applicability to vertical plumes. However, as Turner notes [27] “since the cap merges gradually with the plume behind it, the exact form [of the velocity profile] to be taken for the cap is somewhat arbitrary”, which also implies the role to be played by buoyancy in the formulation is somewhat arbitrary. This has led Abramovich & Solan [28] to successfully apply Turner’s theory to non- buoyant transient laminar jets where the cap is assumed to gain mass solely through addition from the jet, though this may be questioned in that, Hill & Ouellette say of Rizk’s photographs [29] “the head vortex…appears to have suffered little entrainment, and appears almost on the verge of pinching itself off from the preceding flow region.”. Nonetheless, Witze [30] and Rubas et al. [31] have demonstrated the applicability of Turner’s model to impulsively started incompressible turbulent jets. These works support the validity of the vortex-ball model far beyond the case presented originally and hence its suitability for engine relevant injections. Using Turner’s model (with an entrainment constant specified by Ricou & ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 11 Spalding [32] as  where KS=0.32 and the ratio d/Zt is 0.25±0.05 from [29]) to support their scaling arguments, Hill & Ouellette [9] are able to show that jet penetration is given by Eq. 2.2  Eq. 2.2 where Γ is a pure constant- the jet penetration constant- with a value of 3.0±0.1. It is of note that Eq. 2.2 is not directly dependent upon injection pressure. Using this expression with the experimental velocity data of Witze [30] to evaluate the arrival of the jet at a set location, Hill & Ouellette confirm the usefulness of Turner’s simple model even though the Reynolds numbers for the experimental data used are three to seven times lower than the suggested lower limit [32] of 30,000 for a turbulent jet. Using flow visualisation and laser Doppler velocimetry (LDV) Cossali et al. [33] have investigated the early stages of injection prior to the formation of self-similarity.  If the injection duration is short enough, or in the initial stages of a longer injection before self-similarity is obtained, the velocity field around the vortex cap is markedly different to that of the quasi-steady-state discussed previously. The jet head does indeed entrain air upon ejection from the nozzle, though it is posited that this is due to wrinkling of the jet boundary and not velocity differences. It is this which causes the initial jet volume to be larger than the final quasi- steady jet. This ‘enhanced mixing’ in the early stages curtails the applicability of the widespread assumption used in applying Turner’s model, viz. that all the vortex-cap entrainment is from the jet. Cossali et al. do not, however, provide an alternative explanation. Of significance is that steady-state conditions are reached on the jet centre-line about twice as fast as for the edge “where the mixing mechanism requires longer times to reach equilibrium.” [33]. This is attributed by the authors to the dominance of inertial effects on the jet axis in the near field; whereas the boundary is primarily influenced by turbulent diffusion of mass and momentum, with correspondingly longer characteristic times, and where the fluctuations may be of the order of the jet half-width [34]. The importance of r.m.s. fluctuations cannot be overlooked, with [35] showing that turbulent mass transport accounts for ~8% in jets.  The hyperbolic decay of mean (or centreline) velocity could reasonably be expected to apply to the scalar field of a jet, and early work by Hinze [11] and Field [36] confirmed this basic premise, with concentration decay from the source given by Eq. 2.3:  Eq. 2.3 where the co-ordinate z is measured from the virtual origin, and the term in parentheses is termed z*. The data of Birch et al. [37] suggest kjc=4.0 with the virtual origin 5.8d upstream of the nozzle exit. Birch et al. report kjc to be in the range four to six, with more recent studies settling on values close to five, though without conclusive agreement (e.g. 5.4 [38], 5.37 [24] & 4.96 [39]). The location of the virtual origin suffers a wide range of reported values when based on velocity since the location at which self-▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 12 similarity is attained varies widely in the literature (e.g. z* ~ 10 [40], 10-15 [29], 25 [37, 41], 30 [42], 50 [24, 43] and 70 [15, 44]). Birch et al. note that, as with velocity, there is an increase in the rate of spread for concentration in the near field before it settles to its (steady) far-field value at z*~10-30. The difference in the velocity and scalar fields is attributable to the fact that turbulence and buoyancy affect velocity and mixing very differently. Using Raman spectroscopy of a methane jet in air, Birch et al. [37] report that the unmixedness  tends to 28.5% in the far field (z*>70).  The initial development of unmixedness is rapid within the first 10de and is attributed to the r.m.s concentration reaching similarity before the mean concentration. The ratio of the concentration (or buoyancy) to velocity spread rates, λ (= Pr-½ [21]) is important to the analysis of the jet, as is the Richardson number  where lm is the characteristic lengthscale of buoyant jets (or plumes), lm=M0 ¾/B0 ½ [39]. It has been found consistently that the radial concentration profile is wider than the velocity profile at the same location with λ~1.19 [24]. As noted by Carazzo et al. [21] “the route to self similarity is different for jets and plumes…[where] self- similarity occurs earlier in pure plumes than in jets [and] the effective turbulent Prandtl number is smaller in jets than plumes.” Fundamental work by Wang & Law [41] using particle image velocimetry and laser induced fluorescence (LIF) suggest that a jet exists for z/lm<0.6 and a plume for z/lm>6 with a smooth transition between the two regimes ([24] find these values to be 1 and 5 respectively). Papanicolaou & List [24] suggest, that a lack of care in assessing the regime of the flow by various experimenters can explain the discrepancy in the constants reported in the literature.    Figure 2-2 Normalised Radial r.m.s. Profiles of Axial Velocity (w), Radial Velocity (u) and Concentration (c),  after curve fit coefficients presented in [41].  ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 13 Typical r.m.s. velocity and concentration profiles are shown in Figure 2-2, though the extent of the intensities reported by different researchers are inconsistent (cf. Table 2-1). Common to almost all is that, in jets, the concentration fluctuations are smaller than the velocity fluctuations.       Wang & Law [41] 0.224 0.27 0.19 Papanicolaou & List [24] 0.22 0.25 0.17 Birch et al. [37] 0.285 0.27 0.32 Table 2-1 Centre-line Root Mean Square Turbulent Fluctuation Values for Jet Concentration, Axial Velocity and Radial Velocity.  Despite their proximity, Papanicolaou & List and Wang & Law draw markedly different conclusions from their findings. Papanicolaou & List [24] assert from their LDA and LIF measurements that buoyancy produced turbulence accounted for twice the transport of turbulence than jet driven turbulence, with an increase in normalised momentum  of ~12% compared to the initial specific momentum flux (though the exact increase is related to the axial motion of the externally entrained fluid). Wang & Law, on the other hand, point to the fact that the normal and shear turbulent intensities in their jet and plume results are so similar that buoyancy has no effect on the turbulent velocity fluctuations, and acts mainly through the mean velocity. Shabbir and George’s work [40] on the vertical plume suggests that the mean energy advection is primarily balanced by the radial turbulent transport, while the radial turbulent transport is responsible for balancing the mean advected momentum and buoyancy forces. This supports Wang & Law in that although the direct effect of buoyancy upon turbulence is substantial it is still small compared to the effect of shear. Buoyancy is the main driving force in the mean velocity field, which in turn produces the shear to enhance the turbulent flow properties [40]. Regardless, there is a high level of self-similarity in the mean and turbulent velocity fields. The same cannot be said of the respective concentration fields however. The peak r.m.s. concentrations in a jet may be 1.65 times that of the centreline mean. The importance of these fluctuations is further built on when concentration minima are considered [24]. Jets may see r.m.s. concentrations as low as 20% of centreline means, with this value dropping to zero beyond η~0.1 ( ). Root mean square concentration (cf. Table 2-2) and velocity (cf. Table 2-1) values tend to decay like their mean counterparts [24] such that for a jet  and , while the contributions to scalar transport from the turbulent fluctuations are often overlooked, Wang & Law [41] have shown that they account for ~7-12%.   ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 14   at η Wang & Law [41] 0.21 0.07 Papanicolaou & List [24] 0.25 0.1 Birch et al. [37] 0.22 0.07 Table 2-2 Maximum Root Mean Square Turbulent Fluctuation Values for Concentration  (with Radial Position) 2▪2▪2 UNDER-EXPANDED COMPRESSIBLE JETS The discussions presented above pertain to incompressible flows, yet at the pressures used in direct injection systems the resultant jet cannot be described as so. Early work by Kleinstein [45] showed that analytical solutions to the compressible laminar jet were tractable, and that the turbulent case could also be tackled [46]. Assuming: negligible axial pressure gradient, Pr~1, Sc~1, Le~1, and using adaptations of Prandtl’s momentum transfer theory and Taylor’s vorticity transport theory (both with an enhanced eddy viscosity that takes into account the thermodynamics of the problem), Kleinstein [46] showed that the axial decay of velocity and concentration (and entropy) follow the form of Eq. 2.4  Eq. 2.4 where κ is the decay constant (0.074 for velocity and 0.104 for mass fraction [46]).  Birch et al. [42] note of Kleinstein’s solution that the agreement in the near field is only marginally better than the hyperbolic formulation (Eqs.2.1& 2.3) and at large distances may be closely approximated by a hyperbola. However, Kleinstein’s formulation is important in that it is shown to agree well with similar incompressible work by Schlichting [47] up to Ma~2.5, which may explain its persistent wide use [31]. The suggested similarity would imply that despite the well-known structural differences between compressible jets (as detailed by Shapiro [48]) and their incompressible counterparts, their inherent behaviour is the same. This is indeed shown to be the case, with the details of the early development region (where any shocks form) and the potential core ostensibly being ‘forgotten’ downstream. It would appear that the work of George [20] may bring this in to some doubt, but for engine related jets the more complex formulation of George provides no significant benefit. An under-expanded jet undergoes a rapid expansion upon leaving the nozzle to equilibrate in pressure with its surroundings. This expansion causes the subsequent jet to act as though its source were much larger than in reality. This discrepancy is easily overcome by specifying a pseudo-diameter- the diameter that would flow the same amount of mass were it subsonic, and is shown by Birch et al. [49] to be given by Eq. 2.5  Eq. 2.5 In later work [42] the pressure ratio was shown not only to affect the pseudo-diameter, but also the location of the jet’s virtual origin. With this in mind Birch et al. [42] proposed the use of a pseudo-source ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 15 to fully describe the jet behaviour. This source would have a diameter given by Eq. 2.5 with a (subsonic) pseudo-velocity at the source to maintain the mass flow rate for this larger diameter. Hill & Ouellette [9] are able to show that the work of Birch et al. is well defined, but that the more simple scaling of Thring & Newby [16] provides results which match equally as well. This has the implication that Eq. 2.2 for jet penetration applies equally well across the sub-sonic and highly underexpanded jet regimes; with the preservation of the jet penetration constant at 3.0±0.1, and noting that although stagnation pressure losses are significant, the ¼ power dependence of penetration on momentum flux minimises the influence of the loss.  Gaseous jets have relatively simple analytical solutions if steady and laminar, and even a transition to turbulent gets allows a certain amount of analytical tractability. However, when considering jets in more detail and their transients it is apparent that entrainment is the mechanism driving their development and entrainment itself is non-trivial to predict. 2▪3 NATURAL GAS ENGINES 2▪3▪1 NATURAL GAS COMBUSTION ENGINES McTaggart-Cowen et al. [50] offer an excellent review of on-road natural gas engines in terms of their market position and some of the larger technical and infrastructure concerns, yet here our attention is turned to the macroscopic combustion behaviour of natural gas engines.  Natural gas engines come in SI, CI and HCCI variants, where the gas may comprise the major or minor stake in the fuel mixture. Historic development has followed the order given, for ease of implementation, with the latest engines offering exceptionally low emissions, high thermal efficiencies and high power.  In directly injected engines the fuel may be burnt as a premixed or non-premixed mixture, depending upon the time and duration of the injection event. Early injection affords the time for enhanced mixing, both diffusive and turbulent, to yield increased homogeneity. Late injection, with shorter times to ignition will yield a more poorly mixed, or stratified, charge which may auto-ignite or necessitate an ignition source. With the partially stratified charge approach used in the current system, combustion may be either pre- mixed or non-premixed depending upon the timing of the fuelling events. Non-premixed combustion is responsible for the burn in the partially stratified charge and direct injection plumes; while if there is a homogeneous bulk charge present, ignition by the PSC results in a premixed flame. The slow flame speed of premixed natural gas combustion results in slow heat release rates, HRR (that are proportional to propagation velocity), which in turn cause combustion instability and low combustion quality. The result is misfire and high UHC emissions, which are even more pronounced (as are cyclic variations), when leaning the mixture to reduce NOx output. Leaning the charge mixture (for any fuel) yields lower NOx ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 16 emissions through the provision of a heat sink from the excess air, which reduces combustion temperatures and hence thermal NOx. Gupta et al. [51] report a low NOx plateau of 1.12g/kWh (~220ppm) for Φ<0.7. Compared to gasoline, natural gas reduces the reactive hydrocarbons emitted when burnt close to stoichiometric, though, with natural gas, total HC emissions do tend to be higher when measured against the same baseline [52] since methane is considered a relatively unreactive compound, though has a significant global warming potential.  Higher charge pressures retard flame propagation and lower the auto-ignition temperature, which although better for HCCI designs increases the propensity for knocking in SI engines. Retarded timing tends to raise the mixture temperature in the end-zone above the knock temperature limit (which is independent of fuel composition [53]).  Further, a drop in combustion duration, Δθc in Figure 2-3, reduces the thermal efficiency and, hence, power, since there is less heat release near TDC.   Figure 2-3 Power Reduction From Increased Burn Duration and Necessitated Spark Advance, reproduced from [54].  To circumvent the slow heat release rate, fuel additives are often used to increase the flame propagation velocity and get it close to the gas maximum of ~0.2m/s (for lean mixtures), which is independent of fuel composition (Payman in [53]). Diesel pilot ignition is often used to increase the strength of the initial combustion, which is dependent upon mixture temperature, jet mixing and jet penetration. This strong ignition encourages a faster second stage burn in the bulk charge with an overall HRR closely related to the percentage of the total charge which was pilot injected. Pre-chambers offer similar benefits, though their design is more complex with a wider range of variables affecting the second stage combustion of the bulk charge, particularly for multi-stage chamber designs [55]. Spark ignition of lean gas mixtures is inhibited by the high spark energies required, though early gas engine designs used a pre-chamber to ignite a rich mixture allowing lean spark-ignited gas engines to be some of the earliest used. Meyers [56] shows that pre-chamber designs with a volume of 2-3% of the cylinder can offer low emissions with high efficiency. However, Kubesh [57] notes that large pre-chambers (~20%) are required to offer full ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 17 throttless operation, though using a more traditional pre-chamber Goto et al. [58] achieve excess air ratios exceeding two.  More contemporary designs use the diesel pilot to the same effect, though the pilot is the source of the majority of the overall emitted NOx, where NOx emissions are proportional to the percent pilot injection. Regardless, optimal gas engine performance (in terms of efficiency versus emissions) can only be achieved when the gas fuel has at least 80% methane; less than this and the overall fuel octane number is insufficiently low and can cause knocking. The increased reaction rates required for ignition can also be achieved through spark plug shrouding, whereby the electrodes are covered with a shroud, the enclosed volume of which is filled on the compression stroke. The enclosed, more quiescent mixture, is ignited producing strong flame jets, rich in radicals, which then ignite the bulk charge [59]. The enhancement of the radical pool through mini-chambers on the cylinder head has also been numerically investigated as a controlled source of auto-ignition, though is still awaiting experimental trial [60].  The high knock resistance of methane/natural gas (RON ~125-135) allows for the use of higher compression ratios than gasoline SI engines and allows an efficiency increase of up to 6% as a direct result [61]. Figure 2-8 shows schematically how CNG engines can achieve a 25% reduction in CO2 emissions when compared to a gasoline baseline, and an overall emissions reduction of ~20% on a CO2- equivalent basis.  Figure 2-4 GHG Reduction Potential of CNG Engines Against Gasoline Equivalents, reproduced from [61].  Further, the high knock tolerance of NG yields a fuel that is well suited to turbocharging, which in itself lends the ability to downsize and meet the strictest emissions standards [62]. As a fuel, natural gas also offers a better ability to homogenate in the cylinder (if required) and an inherent reduction in nmHC’s. Adaptation of existing engine configurations does not fully access all these thermodynamic advantages however, and there exists the need for bespoke NG engine designs. Bespoke engines could offer cylinder conditions aimed at optimising the burning velocity, not mixing, and hence reduce the required turbulence levels, minimizing wall heat loss. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 18  Gaseous natural gas displaces about 10% of the cylinder air, reducing the volumetric efficiency that, when coupled with the stoichiometry of gas combustion, reduces the quantity of fuel available to produce power by about 15% and hence power by 10-15% [51]; while maximum torque levels are also seen to drop by ~10%. Volumetric efficiency is further reduced due to a lack of charge cooling with gaseous fuels, though this does mean that on cold start no enrichment is required and CO emissions are correspondingly reduced. It is further noted, that on a well-to-wheel basis, CNG shows the ability to offer 15% CO2 savings over gasoline, beneficial but lower than the 25% when measured on a tank-to-wheel basis [63].   2▪3▪2 LEAN BURNING NATURAL GAS ENGINES The stoichiometry of a reaction describes the extent to which there are enough reactants to complete the reaction, which in combustion are fuel and oxidizer. When there is enough oxidizer (or fuel) to complete the reaction with no excess or shortfall, the ratio of fuel to oxidizer, i.e. the fuel air mixture, is said be stoichiometric. Any excess fuel and the mixture is rich, any shortfall and the mixture is lean. The ratio of air to fuel (by mass), AFR, is used to quantify the stoichiometry of the mixture. Mixture composition is often expressed in terms of the relative air fuel ratio (RAFR), λ, which is the ratio of the AFR to that of the stoichiometric condition, the inverse of which is known as equivalence ratio, Φ.  Spark ignited engines must operate at an RAFR of around one, and so load control is obtained by throttling. There are significant pumping losses associated with this which are further exacerbated with natural gas. They range from 10% at full-load to over 25% at part-load, when compared to a diesel engine- a natural gas engine’s obvious competitor [57].  Using a fuel that allows a move away from stoichiometric could potentially reduce, or eliminate, the need for throttling. Natural gas is well suited to such lean-burn applications because of its wide flammability limits and propensity to homogenate (even upon cold start). Modelling of a medium duty throttless naturally aspirated natural gas engine shows that at low load an unthrottled engine requires equivalence ratios as low as ~0.2 [64].  Late intake valve closure can achieve 15% thermal efficiency gains at up to 15% load, with throttling used at part- and full-load [65]. Engine load range can also be controlled by internal trapping and exhaust gas recirculation with supplemental charge heating [66]. It does however, result in higher NOx at stoichiometric conditions compared to gasoline, since a lower residual gas fraction is employed because of significant differences in the required valve overlap. Honda appears to have overcome these problems with an electronically controlled VVT CNG engine with a variable length air intake [67]. Power is reported to increase by 15% over the non-VVT case, with a 5% reduction in fuel consumption.   ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 19 Although methane has a similar ratio of specific heats to gasoline, advantage is to be had in thermal efficiency since as the charge is leaned the mixture ratio of specific heats increases monotonically to that of air. Further, volumetric efficiency gains can be realised over turbocharged and normally aspirated SI engines in that the need for more charge air necessitates a wider throttle with lower pumping losses. Karim et al. [68] and Klimstra & Jones [69] both report natural gas’ ability to operate over a wide range of equivalence ratios with Gupta [51] reporting the same 10% COVIMEP 6 at λ=1.6 compared to 1.4 for gasoline. Yet there are problems with gaseous fuelling and lean operation: high ignition energies and slow combustion durations.  The first of these problems may be overcome with high energy ignition sources such as laser ignition [70, 71], pilot charges or pre-chambers, where laser ignition is also shown to have emissions benefits over traditional spark ignition. The flame speed of a hydrocarbon mixture is strongly related to both the equivalence ratio and pressure of the charge, dropping rapidly with pressure and RAFR’s above one. This is a serious problem for high pressure (large compression ratio) lean burn engines where extinction and misfire are to be avoided.  Hassaneen et al. [72] report the initial flame kernel growth to be retarded by up to 60% at MBT, and the rapid burn duration by up to 33% when a gas mixture is leaned to Φ=0.6 against a stoichiometric baseline. However, the COVIMEP was typically less than 5%, showing good combustion stability. Increased levels of charge motion would elevate burn rates and permit the use of higher compression ratios or boosting, while avoiding knock through extreme leaning.   Figure 2-5 Operating Envelope for an SI Gas Engine,  reproduced from [61].  Figure 2-5 shows, schematically, that with an increase in turbulence, and hence burn rate, the engine operating envelope extends such that higher compression ratios and leaner mixtures can be supported.                                                        6 Coefficient of variation, CoV=σ/µ. r c  λ ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 20 Using this approach Pischinger et al. [61] show the best full load fuel consumption compared to similarly specified SI and boosted CI engines and with 25% less CO2. Combustion chamber and piston design are key to enhanced combustion through turbulence generation. The bowl-in-piston approach is most often used to generate squish and tumble [73-75]. A novel approach outlined by Evans et al. [76-79] uses a fence around the bowl, with gates designed to guide and funnel the squish motion into highly turbulent jets that converge at the ignition location.  Meyers & Kubesh [80] propose a markedly different approach to the use of lean burning engines. Their concept fuels a portion of a multi-cylinder engine with a very rich gas mixture, while the remaining cylinders are fuelled with a lean mixture of natural gas, air and the rich cylinder exhaust. Catalytic treatment of the high HC and CO concentrations in the rich cylinder exhaust is used to produce hydrogen that is then employed to enhance the combustion of the lean cylinders. This approach can offer NOx as low as 8ppm, good combustion quality, combustion stability and thermal efficiencies from 24-28%.  Homogeneous charge compression ignition of lean gas mixtures is reported as pre-mixed charge compression ignition by Kawasaki et al. [81]. It shows that maximum specific power may be increased with engine speeds up to 2400rpm while maintaining an indicated thermal efficiency of 32% and NOx emissions below 100ppm. However, “an increase in the engine speed extends the combustion duration, especially under a lean condition, and so the oxidation reaction is frozen at the last stage of combustion, and the indicated thermal efficiency decreases compared to the low speed condition.”  2▪3▪3 STRATIFICATION AND THE PARTIALLY STRATIFIED CHARGE The aim of stratification is to combine the benefits of a spark ignited engine and a compression ignition engine in a single platform.  The inducted charge is globally beyond the lean limit, but through judicious charge motion the portion of the charge which is ignitable is brought to the sparkplug for ignition.  Stratified charge engines can operate over a wide range of loads through controlled mixture strength, are knock resistant, and can accommodate a wide range of fuel compositions [82]. Abata [83] provides a detailed discussion of traditional stratified charge approaches and their application, with Toyota [84], Honda [85] and Texaco [86] all releasing commercial stratified charge concepts. Zhao et al. [87] offer discussions on the more contemporary approach of direct injection stratified charge (DISC) based on the now widespread practice of gasoline direct injection (GDI). These approaches use a direct injection fuel jet that is wall or piston guided to the plug, during which time entrainment provides the mixing required to form an ignitable mixture. The complexities of the system (fluid-fluid interactions, fluid-surface interactions, injection pressures/timing, piston motion etc.) still remain problematic in that poor fuel utilisation results in high UHC emissions, especially at part load. Further, overmixing and poor delivery to the sparkplug result in significant misfire (with its own emissions penalty) [57, 88, 89], though DISC ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 21 shows the potential for significant efficiency improvements [90-92]. The ignition problems associated with stratified charge may be overcome with local charge enhancement or partial stratification; whereby a small portion (typically <5% w/w) of the fuel is concentrated in the region of the ignition source. This easily ignitable source then provides the energy required to ignite the remaining fuel. This is particularly important in the ignition and stability of ultra-lean mixtures which require high ignition energies [93, 94].   Partial stratification is known to enhance the combustibility of a mixture through increased turbulence generation; Acroumanis et al. [95] report an increase in peak cylinder pressure of 55% and similar flame speeds at equivalence ratios 0.2 lower. At the University of British Columbia a partially stratified charge (PSC) approach has been developed by Evans [7] whereby the pilot charge is injected next to the spark electrodes using a modified spark plug [8, 96]. The plug is then able to ignite the fuel rich pilot which can inflame the rest of the homogeneous charge. Extensive testing has been carried out on a fully instrumented Ricardo Hydra single cylinder engine, the details of which are described elsewhere [8, 96, 97]. No benefit in emissions or engine performance is found for λ<1.4,  although when λ>1.6 at full load, there is a significant extension of the lean limit with an increase of 7% in BSFC. Of more significance are the part load improvements. As is clear from Figure 2-6 there are gains to be made in BSFC and NOx emissions, which are coupled with (not shown) marked combustion stability improvements.  Figure 2-6 Lean Limit Extension & BSFC Reduction (left) and NOx Reduction Potential (right) of PSC reproduced from [98].  The work of Reynolds [96] and Brown [97] (who completed his studies on natural gas PSC with port injected gasoline) was built upon by Gorby whose “goal was [to] determine if mixture enrichment local to the spark plug would aid the combustion of a stratified DI fuel charge” [8]. The thinking behind the work was to improve fuel usage through a larger ignition zone for the DI plume, and to provide an ignition which was stronger and which could then progress in areas which would otherwise be overly lean. These benefits should engender a reduction in UHC’s and “increase efficiency from improved combustion and reduced misfire.”. The potential gains are shown schematically in Figure 2-7 as given by Gorby [8]. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 22  Figure 2-7 Fuel Efficiency of Load Control Strategies, reproduced from [8].  These hypothesised gains were not observed, however. The optimal BSFC was found at early injection timings, the combustion performance degrading rapidly past a minimum threshold value. Further, the DI jet was most effective when directly aligned with the PSC injection; showing that only small gains were to be made from the increased volume of the ignition source. Finally, and most significantly, it was found that the PSC charge was not being reliably ignited and required a certain level of background fuelling to ignite, and a pure air charge impeded the ignition to the extent that it would fail to inflame the PSC injection. However, when the PSC charge was ignited reliably (COVIMEP<5%) the late DI also ignited reliably and provided stable operation of the engine.  It is imperative that a strong combustion be caused through vigourous charge motion and charge ignition if natural gas is to be used as an alternative fuel for ICEs.  2▪3▪4 GAS COMPOSITION Gas engines are known to have a sensitivity to gas composition that acts primarily through the observed ignition delay and the fuel/dilutant kinetics [99]. Ethane (the second most prevalent component in natural gas) decreases ignition delay and combustion duration [99, 100], while the CO2 in EGR gas has the reverse effect [99]. However, the type of ignition source has the dominant effect on ignition/composition interaction. In an unassisted (auto) ignition, non-reacting species such as nitrogen, are known to stabilise the ignition, while no benefit is observed when an ignition source such as a pilot charge is used [101]; highlighting the benefits of enhanced ignition sources to account for fuel composition variability. Further, Schiffgens et al. [102] and Pischinger et al. [61] both suggest that no matter, knock-controlled and lambda-controlled operation is required to account for the range of methane numbers (a measure of the composition of natural gas) encountered when fuelling gas engines. With engine combustion affected, delivered power is also affected. Kim et al. [103] suggest that such a clear relation exists between gas ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 23 composition and engine power that a simple relation based on the Wobbe Index7 may be used as a practical method for power estimation based on fuel composition, although composition is not an integral part of the current. 2▪4 POLLUTANTS AND ENGINE EMISSIONS Pollutants include, but are not limited to: oxides of nitrogen (NOx), carbon monoxide (CO), hydrocarbons (HC’s) and particulate matter (PM). Fuel bound sulphur is of concern since it will form SO2 during combustion, which can then oxidize to SO3, the source of acid rain. It is of note that the US EPA has recently recognised carbon dioxide as a pollutant (along with, methane (CH4), nitrous oxide (NO), hydrofluorocarbons (HFC’s), perfluorocarbons (PFC’s) and sulphur hexafluoride (SF6) ) [104].  The term NOx covers both nitric oxide (NO) and nitrogen dioxide (NO2), which in engine emissions total concentrations of 500-1000ppm [82]. Hydrocarbons from unburnt or partially oxidized fuel, blow-by, fuel evaporation or (in older engines) carburettor release may reach 3000ppm. Intrinsic to these releases there may also be HC’s from oil films, quench products, crevice volumes and oil layer scavenging. Carbon monoxide resulting from ICE’s can reach concentrations as high as 1-2%.  NOx levels are similar for both SI and CI engines, while HC’s may be up to five times higher for SI engines as a result of the exhaust process. There is little PM found in SI exhaust, though it may range from 0.2- 0.5% (at 0.1µm diameter) of fuel mass for CI engines. The high pressures and ample quantities of oxygen found in the combustion process of non-premixed DI engines means that there is little CO present compared to that produced by an SI engine as the thermodynamics is driven towards complete combustion and the production of C02.  Detailed chemistry is needed to link pollutant formation due to combustion chemistry to that of the post combustion processes. This need arises since cylinder concentrations can differ significantly from equilibrium values in the exhaust stream. Fuel chemistry tends to drive CO, PM and organic pollutant formation, while post-combustion processes govern NOx and SOx emissions.  High combustion temperatures will form NO that is frozen during the expansion stroke. Rich combustion yields higher CO concentrations since there is not enough oxygen present to complete the fuel oxidation                                                        7 The Wobbe Index, IW, is define as  and allows the direct comparison of different fuels on an energy basis. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 24 and the CO is quenched on expansion. Lean combustion will have similar results though these are attributable to dissociation rather than partial oxidation.   Figure 2-8 Engine Pollutant Concentrations Across the Range of Operating FAR’s,  reproduced from [82].  Emissions for SI engines are almost solely governed by equivalence ratio. With the exception of NOx, pollutant concentrations rise with increased equivalence ratio as combustion quality starts to deteriorate, cf. Figure 2-8. On warm-up, HC and CO emissions increase because of the rich combustion, while under normal loads, exhaust gas recirculation (EGR) drops NOx by lowering the combustion temperature, though at the expense of combustion quality. Emissions for CI engines are almost solely governed by fuel distribution.  2▪5 OXIDES OF NITROGEN Oxides of nitrogen are integral to the formation of tropospheric (low level) ozone, as first noted by Haagen-Smit [105] and discussed extensively since (often with geographically specific chemistry, see for example [106-112]8).  NOx is also the cause of photochemical smog, which is discussed more extensively in §2▪10. The nitrogen required to form NOx can be supplied from the air charge or may be fuel bound. Nitric oxide predominates in NOx and is formed through several mechanisms [113].                                                        8 The first seven hits for most recently published material on the NOx-ozone relationship at the time of writing (July 2009). ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 25 2▪5▪1 NO FORMATION MECHANISMS 2▪5▪1▪1 ZEL’DOVICH (THERMAL) MECHANISM First postulated in 1946 by Zel’Dovich and proven by Baulch et al. in 1994 [114] (with Lavoie et al. [115] establishing the significance of step (c) in Eq. 2.6) the thermal mechanism is the most commonly cited route to NO formation, viz.:    (a) (b) (c) Eq. 2.6 Kuo [116] also suggests that the mechanism include the primary step:  Eq. 2.7 Steps (b) and (c) (of mechanism Eq. 2.6) are strongly temperature dependent and necessitate temperatures in excess of ~1800K to exceed the high activation energy of nitrogen’s triple bond. Reducing reactant concentrations or lowering the temperature will ultimately reduce final NO concentrations. The intrinsic timescales of the formation mean that the process is kinetically controlled and does not result in the equilibrium condition for the exhaust gas state [117]. 2▪5▪1▪2 FENIMORE (PROMPT) MECHANISM The prompt mechanism first proposed by Fenimore [118] has the CH radical as a prerequisite, which limits its applicability to the flame front (cf. Fig 17-4 of [119] for the complex oxidation mechanism of C1 and C2 hydrocarbons to produce the required CH radical). C2H2 is a precursor to CH and is primarily formed from CH3 recombination in rich flames [119], though the mechanism can be significant at temperatures as low as ~1000K.  2▪5▪1▪3 N2O MECHANISM A nitrous oxide route to the formation of NO was suggested by Wolfrum as an alternative to the two major mechanisms discussed above, and is more prevalent in situations where the relative air/fuel ratio is above 1.6 [120]. The mechanism is normally overlooked, but under lean conditions CH concentrations are suppressed (negating the prompt mechanism) and temperatures are low (eliminating the Zel’Dovich route). The nitrous oxide route is prompt at high pressure (as with all third body reactions).   2▪5▪2 NO FORMATION IN SI ENGINES Early combustion contributes significantly to NO which is then frozen on expansion and may contribute many orders-of-magnitude more to the overall NO concentration [82] than late formation NO. The lack of significant bulk motion near the sparkplug also increases NO concentrations in this region. Given that the ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 26 thermal route to NO production predominates in SI engines, it would be expected, and has been shown [121-123], that NO concentrations mirror the temperature gradients across the cylinder. Spark ignition engines achieve maximum combustion temperatures at an equivalence ratio of ~1.1, but the low oxygen concentration at this equivalence limits NO formation. As the charge is ‘leaned out’ the extra oxygen compensates for the drop in temperature such that NO concentrations peak at about Φ=0.9. Under lean conditions NO freezes early so there is little decomposition compared to rich mixtures, meaning that the gas composition at peak pressure is important. 2▪5▪3 NO FORMATION IN CI ENGINES The kinetics of NO formation in CI engines are also primarily driven by the Zel’Dovich mechanism, though in a markedly heterogeneous fuel and temperature field. Some fuel is pre-mixed and burnt near a stoichiometric composition, while the remainder is non-premixed and burnt at stoichiometric. The critical stage for NO formation is, as with SI engines, at peak cylinder temperature, which is between the start of combustion and peak pressure. Early burning gives high NO concentrations that are frozen by subsequent expansion and the mixing of late burn (cooler) gases or air. It is this fact that limits NO decomposition in CI engines, with almost all NO formation taking place within 20CAD of start of combustion [82].   Increases in overall fuel/air ratio increase both NO and NO2 since it generates higher peak pressures and temperatures. Moving towards a rich mixture reduces NO, though the heterogeneity of the fuel distribution means the reduction is less marked than in SI engines. Most fuel is still burnt close to stoichiometric, with the result that NO formation is almost proportional to total fuel mass injected.  2▪6 CARBON MONOXIDE Carbon monoxide (CO) is a known poison and is intrinsic to photochemical smog formation. CO production is primarily controlled by fuel-air-ratio. The higher the FAR for a fuel rich mixture the more CO is produced; while for lean mixtures CO concentration is independent of FAR, with mole fractions of  ~10-3. Since SI engines run at stoichiometric under part load and rich at full load CO emissions are significant. CO levels are observed to be lower than their maximum possible concentrations, but above equilibrium levels, which suggests the CO mechanism is kinetically controlled. Lean conditions promote CO concentrations well below those suggested by a kinetically controlled model, implying the partial oxidation of oil and crevice HC’s during expansion and exhausting.   Reductions in CO levels have, to date, been achieved through more uniform mixture generation, charge leaning, reduced cylinder-to-cylinder variation, after treatment and better fuel metering during transients [82]. Mixture non-uniformity and transients are still the main sources of CO however [124].  ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 27 2▪7 HYDROCARBON EMISSIONS Hydrocarbons (HC’s), sometimes also referred to as tHC’s (total hydrocarbons) or UHC’s (unburnt hydrocarbons) are more strictly organic emissions. They result primarily from incomplete combustion. Total HC emissions are a good proxy for combustion efficiency, though not total pollution emissions [82].  Fuel composition is important to HC emissions in that rich combustion is known to increase aromatic and alkene concentrations, both of which are highly reactive. The combustion process, through pyrolysis and synthesis, changes the hydrocarbon structure significantly from those bound in the fuel. Oxygenates in the exhaust such as carbonyls and phenols (which are increased by the addition of alcohols to fuel) are important to the formation of photochemical smog (cf. §2▪10) 2▪7▪1 SPARK IGNITION DERIVED HYDROCARBONS Typical levels range from 1000-3000 ppm C1 for SI engine emissions. Hydrocarbon emissions rise significantly on the rich side of stoichiometric (cf. Figure 2-8). The four processes primarily at work are identified by Heywood [82] as: flame quenching, crevice volumes storage, absorption/desorption into/out of the oil layer and incomplete combustion (from high EGR, poor spark timing or mixture leaning). Maximum HC generation occurs on blowdown and at the end of the exhaust stroke with a 50/50 split on a mass basis (though the blowndown HC’s tend to be significantly heavier) [82]. 2▪7▪1▪1 OVERLEANING & OVERMIXING Upon injection a fuel spray develops a mixture distribution similar to that shown schematically in Figure 2-9. The amount of fuel mixed beyond the lean limit (Φ=0.3) rapidly increases with time [125]. The auto- ignition location is normally on the lean side of the stoichiometric contour, downstream of the leading edge of the jet. At this location there should be fuel which has spent the longest time in the combustible limit. The overmixed fuel will not support auto-ignition or a fast reaction, and is only capable of sustaining thermal oxidation, which is most likely to be incomplete. The results are unburnt fuel, decomposed or partial burnt products, most of which will be exhausted from the cylinder. The quantity exhausted will depend upon the amount of fuel injected during the ignition delay [126] and the cylinder conditions at ignition.  ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 28  Figure 2-9 Fuel Jet Structure in a Compression Ignition Engine,  reproduced from [82]. 2▪7▪1▪2 UNDERMIXING Undermixing is attributable to: low injection velocities, i.e. fuel injected late in the cycle, and that which escapes the sac volume; as well as excess fuel from overfuelling, i.e. poor load matching and metering. Fuel retained in the sac volume is vapourised as the cylinder temperature increases and enters the cylinder slowly, and may miss the primary combustion altogether. Heavier fuel fractions may stay in the sac and later be evacuated as UHC’s or undergo post-combustion oxidation in the elevated temperature prior to blowdown. It has been shown [127] that the fuel retained, not in the sac, but in the holes of the injector, matter to overall HC emissions for CI engines. Direct injection compression ignition engines are limited to Φ~0.7 at full load to avoid smoking. They are lean overall, but locally rich during transients due to overfuelling in the power stroke. Hydrocarbon pollutant levels are ostensibly constant for increasing equivalence ratio (at constant speed and minimum ignition delay) until Φ=0.9, when there is a sharp jump in emissions, cf. Figure 2-10. This is critical to emissions under acceleration, although levels are still lower than those observed from overleaning [125].  Figure 2-10 The Effect of Overfuelling on Exhaust HC Concentrations,  reproduced from [82]. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 29 2▪8 PARTICULATE MATTER Particulate matter (PM) is now considered the most important pollutant with respect to human health. The side effects of ingesting and breathing PM can be profound and long-lasting. In SI engines particulate matter can form from lead, sulphates and organic matter in the fuel, though the phasing out of leaded fuel has all but eliminated PM from lead compounds. Organic PM is mostly soot, while sulphates in unleaded fuel (at levels of 150-600ppm) are oxidized to SO2 and then SO3, which reacts with atmospheric water to produce sulphuric acid aerosols. Unleaded fuel generates PM emissions at ~20 mg/km comprised of mostly soluble and condensed organic matter. These emissions result from poorly adjusted engines and, under rich running, may contain significant amounts of black carbon (soot). Compression ignition engines emit mostly carbonaceous material with some adsorbed organic compounds, both of which originate mainly from incomplete combustion of fuel hydrocarbons.   Heywood [82] provides and excellent summary on the formation and nature of PM, where he points out that the composition of CI engine emitted PM is strongly affected not by the fuel make-up but by the exhaust system, and somewhat paradoxically by the analysis collection system. At temperatures above ~5000C individual spherules with diameters of 15-30nm are observed whilst below this temperature the spherules are generally coated, despite their high porosity, in condensed material. The condensates may be any one or a combination of UHC’s, oxygenated HC’s (primarily ketones, esters and ethenes), PAH’s or some inorganic species such as SO2 or NO2. Fuel oil has been shown to contribute between two and 25% of exhaust particulate matter and as much as 80% of the organic material. Traces of zinc, sulphur, calcium, iron, chromium and potassium in PM have also been found with the zinc, calcium and chromium linked to engine lubricating oil. 2▪8▪1 FORMATION PROCESSES Soot originates primarily from fuel-bound carbon, that is to say chains of about 12-22 carbon atoms at a hydrogen to carbon ratio of about 2:1. This yields about 105 carbon atoms with an H:C ratio closer to 0.1:1 [82]. The processes at work to make this transformation are poorly and incompletely understood. The soot formation process, which has characteristic timescales in the order of milliseconds, is known to require high temperatures (1000-2500K) with pressures from 50-100atm and overall enough air to completely oxidize the charge.  Particulate formation requires the production of a condensed phase, most often from fuel species via oxidation or pyrolysis. Generally this phase consists of unsaturated hydrocarbons and PAH’s, which undergo condensation reactions to form the soot nuclei (diameter less than ~2nm). Large numbers of these particles exist, though they form a negligible amount of the total sooting because of their small size. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 30  Particle growth occurs according to: formation, surface growth, coagulation and agglomeration. Surface growth is primarily solid phase through gas-phase deposition, condensation and solidification, and is associated with dehydrogenation [82].  Oxidation may occur at any of the growth stages to yield CO or CO2, and soot formation is an inherent balance between formation and burn-out. 2▪8▪2 FORMATION LOCATION In-cylinder particle distributions are almost solely related to fuel distribution and local heat release. Direct injection results in peak PM concentrations (mostly soot and hydrocarbons) along the jet centre-line. This locally rich core ‘converts’ up to 50% of the local fuel carbon to PM, with fuel pyrolysis in this core contributing the majority of the cylinder-formed soot. The concentration of particulate matter drops away from the jet centre-line, though near the outer edge of the jet, or in the vicinity of the piston bowl, PM concentrations again rise, albeit to levels an order-of-magnitude lower than the core. 90% of all soot is formed in the cylinder prior to exhaust [82]. Particulate formation increases significantly at about four or five degrees ATDC with peak number density occurring at ~200 ATDC. There is then a rapid drop because of coagulation and (possible) oxidation. The volume fraction of PM is seen to rise smoothly much like the number density; however it peaks earlier at ~180 ATDC and drops steadily until oxidation stops at ~400 ATDC, when the PM concentration remains constant [82]. 2▪9 HYDROCARBON COMBUSTION & CARBON DIOXIDE PRODUCTION Water and carbon dioxide are the lowest energy states of the hydrogen-oxygen and carbon-oxygen molecular pairs, and as a result it is inevitable that these will be the most prevalent products of hydrocarbon combustion. Aliphatic fuel combustion is essentially a series of fragmentations in which the chain is broken down into subsequently smaller and smaller intermediate species, each of which is oxidized according to the strict hierarchy of Figure 2-11.   Figure 2-11 Hierarchical Nature of HC Combustion,  Reproduced from Westbrook [128]. ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 31  The intermediate species CO and H2 are common to all parent fuels as are the radicals H, O, OH, HO2 and HCO [116]. The CO is slowly oxidized to CO2 while the other species are key to the oxidation of hydrogen. The detailed kinetics of HC combustion are elucidated most comprehensively by Westbrook and Dryer [128, 129].  2▪10 PHOTOCHEMICAL SMOG & TROPOSPHERIC OZONE As has been alluded to, engine emissions, primarily NOx and UHC’s, but also CO, provide the precursors to photochemical smog. Photochemical smog has been shown to consist of generic hydrocarbons, peroxyacetyl nitrate (PAN, C2H3NO5), NOx, ozone and nitric acid (NHO3) [130]. Tropospheric ozone (with desirable levels below 0.1ppm [131]) is highly stable and its measurement thus acts as a proxy for general air quality. ‘Difficulty regulating O3 occurs because in regions of high NOx (primarily urban centres and power plant plumes), O3 formation is limited by the availability of hydrocarbons.’[132], while in rural regions it is the availability of NOx which is the limiting factor in O3 production [133].  The addition of VOC’s (most commonly biogenic isoprene [132] ) and CO to the atmospheric chemistry alters the reaction mechanism significantly, since the VOC’s initiate different precursor reactions [134]. The complex interactions of the O3-NOx-VOC system ensure that smog chemistry is strongly non-linear and the concerned reader is referred to Sillman’s seminal paper The Relation Between Ozone, NOx, and Hydrocarbons in Urban and Polluted Rural Environments [134]. 2▪11  HEALTH IMPLICATIONS OF ENGINE EMISSIONS The pollutants discussed previously are significant in that their role on atmospheric chemistry and conditions may be contended, but the negative effects of engine emissions on respiratory health are beyond doubt. Hydrocarbon volatiles and aldehydes irritate the eyes and respiratory tract, while phenols have a similar effect and are an odorant. Hydrocarbons, as has been noted, are critical in photochemical smog formation, which is itself an eye and respiratory tract irritant. Aromatic compounds resealed from CI engines are known carcinogens, while contemporary thinking puts PM at the fore of health concerns related to air quality. Tropospheric ozone is know to cause significant respiratory morbidity and has been related to mortality [135, 136]. The problem of urban air quality management is difficult in that the chemistry is non-linear and temperature dependent [137].   The PM problem is complex. There is ‘emerging evidence of PM-related cardiovascular health effects and growing knowledge regarding interconnected general pathological pathways that link PM exposure with cardiopulmonary morbidity and mortality’ [138]. Despite toxicological studies of the effects of PM upon ▪ Chapter 2 ▪  Gaseous Jets, Gas Engines & Pollutant Emissions ▪ 32 respiratory health [139-143] establishing a direct casuality between PM and mortality/ill-health is not without significant problems, ‘however, recent research has increased confidence that the PM- cardiopulmonary health effects observed in the epidemiology are “biologically plausible”.’ [138]. The concentration-response function for PM is almost conclusively linear and as such ‘further improvements in air quality are likely to result in corresponding improvements in public health’ [138]. There is still a significant lack of knowledge, however, in our understanding of which pollutants, or combinations thereof, cause which symptoms and the role of coarse and ultra-fine PM in the pathological systems.  2▪12 CLIMATIC IMPLICATIONS OF ENGINE EMISSIONS The realm of climate change is vast and nebulous. The discussion about it in the scientific and popular communities has been vitriolic and heated. The Earth’s atmosphere contains a large number of chemical species, some natural, and many anthropogenic. The role of these chemicals is wide and varied, but even small concentrations play a significant part in regulating tellurian climate.   The complexity of the climate system makes it almost impossible to model. Progress has been steady as atmospheric models have started to include more detailed chemistry and the role of water vapour, though the role of the complete hydrological and carbon cycles are not fully understood and are not at the stage of being accurately modelled. Many of the climate sceptics’ main arguments for non-anthropogenic forcing (solar activity, poor data and temperature lagging CO2 increases) have been recently debunked. As the UN IPCC makes clear in its latest report [144] much of the data pertaining to climate change is incomplete but the science is still sound and must act as a foundation for future action. Current atmospheric CO2 levels are around 380ppm, and the IPCC reckons that stabilising these at ~540ppm is economically viable and ‘safe’. Any increase beyond this will significantly endanger human activity on the planet. The IPCC report [144] predicts temperature rises anywhere from 1.1-6.40C in the next 100 years. A clear ‘bottom line’ has been drawn under anthropogenic radiative forcing. Regardless of the benefits to human health, reductions in the level of pollutants emitted from ICE’s is imperative if the climate change problem is to be mitigated. 2▪13 CONCLUSIONS Arguments have been presented that describe the behaviour of incompressible and compressible gaseous jets. The suitability of natural gas fuelling for transport engines has been argued and a description of the main technological approaches used was made, with particular attention paid to partial stratification- the technology at the core of the current work. The nature of the pollutant formation in engines has been discussed. The general nature of hydrocarbon combustion has been offered before considerations were made to pollutant effects upon human health and the climatic environment.    33 Chapter 3  LASER IMAGING TECHNIQUES & PLIF FUNDAMENTALS   3▪1 INTRODUCTION Chapter 3 provides an overview of several techniques used in laser-based species concentration measurement. The Chapter provides the most relevant details of each approach, and briefly outlines their suitability for combustion diagnostics. The reader is taken through the approaches offered by: Rayleigh and Mie scattering; Raman spectroscopy; advanced Raman techniques; and laser induced fluorescence. The majority of the Chapter is concerned with the basic principles of planar laser induced fluorescence (PLIF), its background and applicability to experimental flows. Building upon this, a qualitative explanation is made with regard to how PLIF can be used in elucidating mixing processes and, in particular, mixture formation in internal combustion engines. A case is made for the use of acetone over the many other possible options as a gaseous fuel tracer, and a detailed explanation of the photophysical process behind acetone LIF provided. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 34 3▪2 BACKGROUND Laser diagnosis provides the ability to investigate harsh environments, such as those in the combustion chamber of internal combustion engines, whilst being non-intrusive and can yield high spatial and temporal resolution. Fluid and chemical parameters can be simultaneously or individually resolved at any point or across a plane, with contemporary techniques providing full three dimensional information. State specific, species specific and non-equilibrium chemical compositions may all be scrutinized and, if desired, related to the temperature field.   Modern lasers have advanced the field of combustion diagnostics through their high powers, coherent light production, spectral purity, short pulse durations (down to ~10ns) and small probe volumes [145]. High laser powers allow the exploration of weak events, including some that were previously unattainable with arc and flash lamps. Further, the coherence of laser light opens up the opportunity to use spectroscopic techniques only available with coherent stimulation. Species and state resolution have been enhanced with the improved spectral purity of lasers and allow specific electronic states to be targeted. The ability to sweep a range of spectra or multiplex multiple light sources presents the opportunity to simultaneously investigate a range of species. The latter point is complimented by the ability to focus laser light sources into probe volumes as small as 50µm2, which, coupled with short pulse durations, allows both chemistry and flow to be frozen on the smallest of scales.  Laser diagnostics can be split into four main areas: velocity determination (both point and plane); the use of Rayleigh scattering for density determination; Raman techniques for major species (~1% concentration [146]); and  laser induced fluorescence for minor species (in the order of ppm).  3▪3 COHERENT AND INCOHERENT PROCESSES An incoherent process will result in the emission of light which is non-directional, that is to say it will be emitted over the full 4πrs steridian solid angle of the molecule/particle. A scattered signal will result from each point along the path of the laser beam and as such allows for easy extension of these techniques to 2D measurement. Coherent phenomena emit light in a specific direction much like a laser beam. The probe volume is now governed by the intersection volume of the incident laser beams, where all the light emitted from the molecule is collected by the collection optics. This gives a stronger signal than incoherent approaches, but makes the experimental set-up more complex since the exact direction of emission must be established a priori. Eckbreth [145] lists the advantages and disadvantages of each as:   ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 35  Incoherent Coherent Advantages Linear Intensity independent Single ported Spectral simplicity Simple calibration Strong signals Laser-like signal beams Interference tolerant Disadvantages Large solid angles Prone to interferences Non-linear Intensity independent Multi-ported Complicated Spectra Difficult to normalize Refraction sensitive Table 3-1 Advantages and Disadvantages to Coherent and Incoherent Laser Diagnostics Techniques,  after [145] 3▪4 VELOCITY TECHNIQUES The importance of the flowfield within a combusting environment cannot be understated and as a result a brief outline of laser based and optical velocity tracking techniques is presented herein. For a full discussion the reader is referred to the excellent books of Durst et al. [147] and Taylor [148] and the review of Adrian [149].  Non-intrusive velocity measurement is now possible, with high resolution, at a single point or across a plane. Particle image velocimetry, PIV, uses a CCD camera to cross-correlate, across a number of sub- regions, laser light which has been elastically scattered by particles seeded into the flow. High laser powers provide a strong signal that can also be enhanced with high seed densities (which are tolerable since particles are not tracked individually but throughout the sub-region of the image). Laser Doppler velocimitry, LDV, uses the Doppler shift in the light elastically scattered from seed particles to ascertain the flow velocity at a point. 3▪5 RAYLEIGH AND MIE SCATTERING Both Rayleigh and Mie scattering involve the elastic scattering of light incident upon a particle or molecule. There is no energy transfer between the incident photons and the target molecule, which means that the scattered light is not frequency shifted. The process is incoherent and for d/λ<<1 is termed Rayleigh scattering while for πd/λ≥1 it is Mie scattering. The lack of photonic interactions means that both processes are species independent and as a result can only provide generic density information and only under some circumstances species concentration [145]. The processes do provide high signal intensity, yet may be prone to spurious light interference. Combination with LDV provides a powerful diagnostic tool [150]. A comprehensive summary of Rayleigh scattering’s use in combustion diagnostics has been presented by Zhao and Hiroyasu [151]. The Mie scattering effect generates a weak signal in ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 36 unseeded flows and is most commonly used in naturally seeded flows, i.e. soot studies where the technique is now in competition with laser induced incandescence (LII). 3▪6 RAMAN SCATTERING Raman processes are characterised by inelastic scattering, and may be based on changes in one, two or all of the rotational, vibrational or electronic states of the molecule. The process is almost instantaneous (~10-12s), hence linear Raman spectroscopy is referred to as spontaneous Raman spectroscopy, SRS. The photons generated from Raman scattering are shifted from the incident frequency by a quantity Δυ, where υ is wavenumber. For Δυ<0 the spectral lines produced are termed Stokes lines while for Δυ>0 they are anti-Stokes (cf. Figure 3-1). The latter requires energy addition from the target molecule and as such the molecule must first be in an excited state or at elevated temperature.  Given the wide range of transitions that can take place to produce a Raman shift the hyperfine structure is grouped according to changes in rotational quantum number, J. Q-branch transitions are those for ΔJ=0 and Δv=±1 (i.e. the principall quantum number changes by one). When ΔJ=±1 the branches are termed P and R-branches respectively, while for ΔJ=±2 the O and S-branches are obtained.  Figure 3-1 Ro-vibronic Structure of the Raman Spectrum,  adapted from [152].  Scattering intensity scales with the fourth power of frequency [145], which means that UV diagnostics are preferable. Further, as a result of energy quantization the Raman shift observed is unique to each chemical υS υR υQ υP υO υO υP υQ υR υS Stokes Lines Rayleigh  Line Anti-Stokes  υ0 ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 37 species and proportional to molecular number density. This all means that Raman is well suited to species identification and concentration measurement; however with a scattering cross section of ~10-31cm/sr signals are weak.  Near-resonant Raman uses laser excitation to excite target molecules near an electronic resonance, thereby increasing the scattering cross section by as much as six orders-of-magnitude [145], similar to those of laser induced fluorescence (cf. §3▪8). The difference is that the incident photon is not absorbed by the target molecule; with the result that the scattering is much shorter in duration, and not subject to quenching effects. However, for combustion diagnostics there are few species which can be excited at near resonant frequencies. 3▪6▪1 HYPER-RAMAN TECHNIQUES Hyper-Raman is the term given to any generic Raman emission that results from multi-photon excitation. The resultant emission is coherent though the excitation is weak, leading to a taxing experimental setup. This is often prohibitive, though coherent anti-Raman scattering (CARS, cf. §3▪7) provides the exception to the rule.   Stimulated Raman gain spectroscopy (SRGS) and stimulated Raman loss spectroscopy (SRLS) are induced emissions techniques at the Stokes and anti-Stokes frequencies respectively. The target molecules are excited with both a pump and probe laser. SRGS and SRLS have the advantage that experimentally they are very similar to spontaneous Raman scattering as they are based on a resonant phenomenon to produce the emission and do so without the lineshape complications of other hyper-Raman techniques, while still providing a signal proportional to number density [145]. Nonetheless, beam multiplexing is required to reduce the interrogation time whilst still ensuring a strong signal. This is difficult to achieve and coherent anti-Raman scattering (cf. §3▪7) is favourable under these conditions. Beam steering that results from turbulent fluid motion impedes diagnostic significance and cannot be corrected for. This, and the experimental complexity, mean that SRGS and SRLS are of primary use in the fundamental research of molecular photophysics.  Raman induced Kerr-effect spectroscopy, RIKES [153] is a Raman based effect whereby the target molecule is pumped (by a two colour phase matched system) such that the molecule takes on a polarized rotation at the Raman frequency. This induces an emission at the corresponding Raman shifted frequency and with the corresponding polarization. A polarizer is used to block the pump light, while the resultant emission still has components from the resonant and non-resonant parts of the excitation, the latter of which requires filtering. The difficulties of RIKES arises from poor signal strength due to extensive filtering and polarization of signal beams both of which contribute to a poor signal-to-noise ratio. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 38 Turbulence in the region of interest also introduces significant anisotropy, for which the induced polarization is likely to be weak. This practically limits the technique to high density media, i.e. liquids, and curtails its diagnostic utility [145].   The problems of poor susceptibility (in essence, the ability to excite a molecule) and incomplete probe laser rejection in RIKES are circumvented with optically heterdyned-RIKES (OHD-RIKES). The details of this technique are somewhat nuanced and not of current concern, though the interested reader is referred to works of Eckbreth [145] and Eesley [154].   Photoacoustic Raman spectroscopy, or PARS, is very much like SGRS except that the collection ‘optics’ are now acoustically based and collect the pressure wave that results from the photomolecular de- excitation. This means that the signal collection is free of all electromagnetic contamination. It is possible to detect signals down to 1ppm, though the noise requirements for this, or anything useful, are prohibitive for diagnostic application. 3▪7 COHERENT ANTI-RAMAN SCATTERING Coherent anti-Raman scattering, CARS, is a three or four colour non-linear technique whereby the pump laser, ω1, and the probe laser, ω2 (which is Stokes shifted from the pump), combine to stimulate the emission of a wave at ω3=2ω1 - ω2 (cf. Figure 3-2). Based upon this approach Eckbreth [145] also reports the technique of coherent Stokes-shifted Raman scattering (CSRS) whereby ω2 is now shifted to the anti- Stokes lines instead of the Stokes lines, though this application is much less widely used and is, as yet, unreported for diagnostic purposes.  CARS can exhibit significant constructive and destructive interference because of the non-linear ro- vibronic interactions of the target molecule, though their interpretation is still analytically tractable [145]. The non-resonant part of the Raman signal can be removed through simple polarization techniques, though this significantly affects the quality of the signal such that the experimental set-up quickly becomes shot-noise-limited. Phase matching is not guaranteed automatically, though since gasses (which are the main target for CARS) are dispersionless, ensures that simple overlapping of the pump and probe beams generate phase matching. This does in turn limit the spatial resolution of the apparatus in collinear systems. Higher resolution necessitates the use of crossed-beam approaches and their inherent phase matching complexities. CARS is most commonly employed to excite the vibrational branches of the target molecule (particularly the Q-branch). However, similar techniques to those outlined here can be used to hit the rotational branches of the Raman spectrum [155]. Rotational CARS (RCARS) allows for simultaneous investigation of multiple species [146] and is less prone to carbon poisoning. Further ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 39 advances in CARS ease of use and applicability have been made by Eckbreth et al. [156] and Alden et al. [155].   Figure 3-2 CARS adapted from [146]   The advantages of CARS are that it provides a signal many orders-of-magnitude larger than SRS; its coherent signal allows for complete collection; there is no fluorescent inference since emission is on the anti-Stokes side of the Rayleigh line and is more robust than SRS in harsh environments [145]. CARS is impeded by its limited sensitivity to species concentration and the requirement for concentrations greater than ~0.1%, though this is similar to that required by Raman, and much higher than that tolerable in many LIF techniques. LIF’s inability to cope with high concentrations (due to saturation) means that LIF and CARS may be seen as complimentary, as has been demonstrated by Mokhov & Levinsky [157]. Miles et al. [158] have also extended the idea to their RELIEF (Raman excitation plus laser induced electronic fluorescence) technique which can provide velocity information. Rotational CARS has also been shown to provide significant thermometric data [159]. 3▪8 LASER INDUCED FLUORESCENCE Laser induced fluorescence uses the energy bound in the incident beam to promote the target molecule to a higher energy state. This ‘optical absorption’ is followed by an almost instantaneous relaxation which may, or may not, be radiative. The entire process has a timescale of 10-10-10-4s [145]. Fluorescence is the radiative emission that maintains the spin multiplicity of the ground state, while phosphorescence is that which results from a state of different spin multiplicity. Phosphorescence is characterised by much longer timescales, typically 10-4-10-2s. The radiative emission may result from a number of states whether excited, predissociated or photoionized; while the yield may be linear or saturated depending upon the exact treatment of the subject molecule. Fluorescent emission can be at the incident wavelength (resonance fluorescence) or, more commonly, hypsochromically shifted from the incident frequency. Fluorescence gives signal strengths many orders of magnitude higher than Raman based techniques and as a result can be used to image very low concentrations (down to trace or ppm) and is readily expandable to planar analyses. Of most concern to the experimentalist is the strong tendency for excited molecules to be quenched. Quenching refers to any process which deactivates a molecule without directly leading to radiative emission. ω3 ω1 ω2 ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 40 3▪9 APPLICATIONS OF PLANAR LASER INDUCED FLUORESCENCE Laser diagnostics have a wide range of uses applicable across a range of transient [160] and steady flows which may be sub-, or even, super-sonic [161-165] (although the latter has its limitations [161, 166]). Flow properties such as: velocity (using metal atom tracing [167], iodine [168-170] and nitric oxide [171] fluorescence), periodic instability [172]; or the extent of entrainment [173]  are obtained easily. With added experimental and computational complexity more fundamental flow properties may be obtained. These include: levels of vorticity and strain [174]; the dissipation length scales of energy and mixture fraction [175]; small scale turbulent structure identification [176, 177]; conserved scalar behaviour [178]; and heat transfer characteristics [179, 180]. LIF is primarily used to study mixing [181, 182] and the resulting concentration field [160, 183-194], with the works of [173, 180, 193, 195-199] on concentration fields in jets of particular note. The technique of LIF may even be extended to visualise concentrations in three dimensions [199]. Advances in molecular photophysics and the role of temperature within it (cf. §3▪11▪6) have extended LIF application to the temperature imaging for, amongst other things, gaseous flows [169, 184, 186, 200-205] and liquid droplets [206].  Laser induced fluorescence has provided much experimental data on combustion and flames. The fuels studied range from coal [207] and acetylene [208, 209] to the more pertinent: LPG (liquid petroleum gas) [210]; hydrogen [160, 184, 211-213]; natural gas [214] and methane [187, 188, 194, 215-226]. Using LIF, it is now possible to gain insight into sooting [227-231], flame surfaces structure [232] and blowout [233]. Existing engine related laser diagnostics cover: flame kernel growth [234, 235]; swirl [236]; heat release rates [237, 238], pyrolysis [239]; flame position [240] and mixture formation (cf. §3▪9▪1); and are applicable to homogeneous (including homogeneous charge compression ignition [241-243]) and stratified mixtures [244-246] alike.  3▪9▪1 LIF FOR MIXTURE FORMATION AND DISTRIBUTION Early work centred on mixture distribution imaging from Rayleigh scattering [247] or with fibre optic imaging of CH and C2 [248]. The C2 signal provided flame position while the ratio of the CH to C2 signals provided equivalence ratio. Established laser techniques such as Raman scattering have been used to investigate major flame species (CH4, O2, N2, H2O and CO2) as well as minor components (H2) [215], the concentrations of which, when combined with Rayleigh scattered light, can provide temperature information [249]. However, for most minor species (NO, CO and OH) LIF is the experimenter's only recourse. Frank et al. [215] use three Nd:YAG lasers to probe methane/air flames with equivalence ratios of 0.6, 0.7 and 0.8. The species concentration results compare favourably with laminar flame calculations. Acetone emission displays only a small temperature dependence at 266nm excitation. However, Fujikawa et al. [250] in their study of gasoline direct injection stratified mixture formation perform a number of cell ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 41 experiments to identify the variation in LIF behaviour at different temperatures and pressures. An analytical approach is used to calculate the DI jet temperature due to evaporative charge cooling and fuel superheating, with the effects upon the LIF signal then taken into account to produce an accurate temperature corrected fuel-air-ratio (FAR) field. Fujikawa et al. use acetone as the fuel dopate, which is substantially different in nature to the iso-octane fuel, wiyj the inaccuracies are compounded by low laser energies (60mJ/pulse) necessitating high seed concentrations (10% w/w). The concerns of differential evaporation are cursorily addressed by noting the highly elevated cylinder temperatures, though the consequences of differential diffusion, fractional distillation, and the effects of the high seed concentration upon combustion are not discussed. It is subtleties like these that must be addressed since, as is noted by Kraemer et al. [251], two seemingly identical initial mixtures may yield significantly different combustion characteristics. Kraemer et al. use a benzene/TEA9 trace for iso-octane mixture formation in a DI SI ICE. Since the tracer is consumed during the combustion event, flame tracking is also performed. Mixture formation and flame front propagation are further studied by Wolff [252] who also investigates relative fuel density and residual gas distribution using an acetone doped (non-specified) fuel for an in-line four cylinder SI engine with equivalence ratios from 0.3-1. Of more use to gasoline, iso- octane and two phase fuel systems is the work of Ipp et al. [253] who use laser induced exciplex fluorescence of benzene/TEA to track both fuel phases. Their technique (which is validated in the liquid phase by Mie scattering and Raman scattering in the gas phase) allows the specification of probability density functions for the FAR at a series of locations. The same tracer combination was used by the same authors [254] to investigate (with Raman validation) mixture formation in lean (Φ = 0.8) GDI engines where substantial fuel concentration gradients are found. The Raman investigation yielded sharper gradients due to its smaller probe volume, while the utility of completing Beer-Lambert attenuation calculations for incident laser intensity was brought into doubt. TEA also has application to gaseous fuel tracking [255] where it yields qualitative injection development information and can be temperature corrected to provide quantitative FAR values that correspond well to simultaneous Raman. TEA was also shown to track iso-octane and helium well (which was used as a safer proxy for hydrogen in motored studies), with only a ±3% error reported at a relative air fuel ratio of three. However, problems with TEA pyrolysis and retained gas fraction were noted to skew the LIF results under motored conditions. No such problems were encountered by Medaerts et al. [214] when they used toluene to track natural gas and gasoline. Using excitation at 248nm, mixture heterogeneity (for Φ=0.9-1.0) was elucidated. Pre- combustion diagnostics of lean hydrogen mixtures (Φ=0.55) is achieved with an acetone trace in the work of White [256], who simultaneously uses OH* chemiluminescence to investigate the flame structure. A lack of detailed knowledge about the quenching behaviour and kinetics of OH* limits the investigation to the qualitative, which is further complicated by line of sight signal integration and the symmetry of the                                                        9 triethanolamine ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 42 OH* intensity about Φ = 1.  Similar imitations are placed upon the LIF signal due to the low seed densities which result from the limited vapour pressure at the high injection pressures. Despite these shortcomings the importance of injector geometry, injection timing and injection pressure are established for gaseous fuelling. Oh & Bae [257] have similar problems with a weak signal in their LPG LIF study, which uses acetone in a lean stratified mixture, though they do obtain meaningful mixture distributions in the richer zones and good flame front tracking. The work of Kaminski et al. [221] compares well with DNS predictions. The LIF signal, which is calibrated against 1D laminar flame calculations, provides good mixture fraction values (±30%) which are ostensibly temperature independent at the excitation wavelength of 282nm. The effect of highly controlled turbulence (the extent of which is verified by laser Doppler anemometry) is also evident, as are the quenching and surface reaction effects of the electrodes upon ignition. By applying a 2µs delay to one of the CCD flames and capturing OH* chemiluminescence, flame kernel structure was also investigated, while the calculation of the Reynolds and Karlowitz numbers allowed the establishment of flame rate and surface to volume ratio. A judicious use of tracer and excitation sources allowed a temperature map to be established simultaneously with a mixture distribution map. The pioneer of this technique, Marc Thurber, details this approach in [258] whereby he (and his colleagues) employs a method that uses a simple acetone dopate (at either 3% or 9%v/v for a co- flowing jet) excited at 248 and 308nm concurrently. This yields both the temperature and mole fraction fields. The short (30ns) lifetime of the fluorescence freezes the flow while careful concentration adjustment and laser sheet alignment ensure good agreement with the existing literature. A similar excitation scheme was used by Einecke et al. [186] to excite acetone’s linear cousin 3-pentenone and make use of its 10nm hypsochromic shift per 100K increase in temperature (for any change in absorption behaviour is reflected in the emitted light intensity). However application was limited to small loads in a two-stroke engine since lean conditions (Φ=0.62) were required to avoid excessive self quenching. 3▪10 TRACER CHOICES The current work is solely related to stratified natural gas combustion, and as such the discussion herein limits the innumerable set of tracer and fluorescing species to those suitable for gas phase diagnostics. Methane, the major component of natural gas, does not fluoresce under the influence of any known radiation sources. The consequence is that it must be doped with a fluorescent compound which acts as a proxy for the methane. The most important parameters for a tracer are those listed below (which have been paraphrased and adapted from [259] & [260]  Possess a high fluorescent yield at easily available high-power laser wavelengths. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 43  Absorption and excitation of the tracer should still leave the fuel/tracer/air mixture ‘optically thin’10.  A good spectral separation between the absorption and emission wavelengths is required so as to ease the implementation of the excitation and collection optics, but more importantly to avoid fluorescent trapping whereby the emitted light may be absorbed before reaching the detection source.  It is desirable to have the fluorescent intensity independent of bath composition, temperature and pressure, yet strong enough to allow 2D visualisation. The lifetime must also be short enough to freeze the flow.  The tracer must be structurally stable at the temperatures and pressures found in ICE’s (for the duration of the cycle at least).  The tracer should be consumed during the combustion process so as to avoid tracer build up.  It should be inert with respect to the combustion process while tracking the fuel well.  The dopate should have a high vapour pressure to facilitate low seed concentrations (which will also help the preceding point).  Finally, the tracer must be non-toxic. 3▪10▪1 METAL ATOMS Metal atoms are known to have large absorption cross sections with strong emission spectra, yet their production generally requires high temperatures [261]. Seeding levels may be kept low because of their strong emission, hence metal atoms allow the use of low power lasers. Consequently, saturation of the signal is easily achieved, which results in low signal strengths, despite the positive spectral properties.  3▪10▪2 INORGANIC MOLECULES Many non-organic compounds are known to fluoresce when excited by a wide range of photon energies. Some, most notably water [262] and oxygen, display a weak signal that requires low pressures (almost vacuum) or high temperatures to fluoresce. Oxygen, even at moderate pressures (~ 10bar) is rapidly collisionally quenched [263]. Despite this, the work of Lee and Hanson [264] paved the way for excitation in the deep UV (193nm), which provided meaningful O2 concentration and temperature results ([265, 266] respectively).  Nitric oxide (NO) has been used extensively to study mixing, diesel sprays [267] and pollutant formation in flames [268] and combustors [269]. The spectral properties of nitric oxide have been exhaustively                                                        10 The incident radiation should exhibit no appreciable change in properties after passing through the medium/object. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 44 reported in two seminal papers by Bessler et al. [270, 271], and build upon the earlier understanding provided by Di Rosa et al. [272]. With an absorption peak at 225.83nm [272] nitric oxide is easily excited with: the fourth harmonic of an Nd:YAG laser (214.34nm) [273]; a tuneable UV laser [271]; or a pumped dye laser [274, 275]. The collection requirements for the broadband UV emission are not excessively onerous. For accurate results the NO LIF signal must be extensively corrected for: temperature dependencies; differential diffusion; self quenching; the quenching effects of other bath gases; and signal interference from O2 fluorescence [276]. The problems of seeding a fuel with NO can be circumvented by mixing the nitric oxide into the oxidiser stream, potentially allowing the mixture formation to be imaged ‘inversely’ [261], though this technique, to the author’s knowledge, has not been applied quantitatively as yet. Further, substantial problems are also encountered due to the markedly different evaporative and diffusive characteristics of NO when compared to most fuels. This is most easily overcome by causing photodissociation of seeded NO2 [171, 176] or the O2 content of air [277], which provide the precursors to NO. This approach does not, however, allow for substantive 2D mapping and is limited by its short lifetime [261]. All these factors mean that NO LIF is suited to combustion events, rather than the mixing processes involved in engine mixture preparation.   Rather than photodissociate NO2, the compound itself may be used directly as a fluorescent tracer. Nitrogen dioxide has many rovibronic excitation levels, and as such has correspondingly wide absorption and emission spectra. Typically, a frequency doubled Nd:YAG (532nm) laser is used for the excitation [273, 278, 279] although the spectra is broadband from ~250-666nm with a peak at ~357nm [280]. The fluorescent emission is seen to fluctuate significantly between 450 and 760nm, albeit with an increase in intensity towards longer wavelengths. Despite the insight into cyclic variations from Zhao et al. [279] on the applicability of NO2 LIF to engine diagnostics, nitrogen dioxide is of more use in examining reaction kinetics; as has been demonstrated by Cattolica et al. [273]. Yet, it must be noted that, in general, NO2 suffers the same problems as NO for tracking mixture formation in ICEs, and is less spectrally favourable for the same applications.  Sulphur is exclusively utilised for LIF studies as sulphur dioxide (SO2). SO2 displays a clear rovibronic structure with absorption throughout the UV, which strengthens at shorter wavelengths (Greenbough et al. [281] use a mercury lamp to excite over the range 180-390nm). A clear peak is observed in the absorption spectra at ~200nm, which produces “a broad structureless emission with a maximum at around 360nm” [282], though Greenbough et al. later pin-point the maximum to 374nm [281]. With a bathochromic shift in excitation Strickler & Howell note a marked increase in the phosphorescent to fluorescent emission ratio (×4, in the range 289-313nm) [282]. This observation is further corroborated by Mettee [283] who studied fluorescent lifetime from excitation at several  wavelengths (313, 302, 296, 285 & 265nm), and attributed the shift to vibrational relaxation having a more marked effect than quenching, which would in ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 45 turn, explain the reduced fluorescent emission at longer wavelengths. Sulphur dioxide emission is heavily self quenched, the quenching being some twenty times stronger than that for CO2 [281], while it is also noted that oxygen only quenches the singlet emission [281, 282]. Nitrogen is also shown by Rao [284, 285] and Mettee [283] to heavily quench the emission, which correspondingly limits the applicability of SO2 to combustion environments. More pertinently, the general observation that there are significant numbers of quench partners means that emission intensity is severely limited by increases in pressure. Through the use of sulphur doped fuels SO2 LIF is deemed to be of more use in studying exhaust gases [260] or residual gas fraction and its influence on mixture preparation than the mixing process itself.  The body of work to support the OH radical as a fluorescent tracer is vast, though its application to the current study is limited. Generation and subsequent seeding of OH into a flow is impractical due to the short lifetime of the OH radical. Photodissociative production of OH from vibrationally hot water has been reported by Pitz et al. [286] to circumvent the seeding problem, though the approach is spectrally limited. The interested reader is referred to [287] for the details of OH LIF (and CH & C2, which are similar in nature).  Iodine is highly corrosive and toxic, but has nevertheless been used as a fluorescent medium because of its favourable spectral properties. Hiller [288], and references therein, review extensively the use of iodine to track pressure, temperature, density and velocity. Of particular note in the application of these approaches are Kido et al. [173] who use iodine to study gas entrainment in intermittent low speed jets; Lefebvre et al. [169] whose experimental apparatus provides simultaneous pressure, velocity and temperature fields and Lemoine et al. [170, 289] who use iodine LIF in compressible flows. Absorption is broadband in the green (usually with a 514.5nm argon ion laser excitation), though has a weak transition probability due to its spin forbidden nature. The forbidden transition generates a long emission lifetime (0.3-7µs, [290]) which allows for velocity specification. Excitation in the green provides a highly resolved ro-vibronic structure that results from nuclear-molecular spin interactions and yields some 45000 lines between 500 and 650nm [288]. Below 499nm the incident energy is enough to predissociate the iodine [288]. Molecular nitrogen and oxygen are known to be good quench partners though good fluorescent efficiency offsets this somewhat. This efficiency is due, in part, to the large absorption cross-section of iodine, yet this ‘advantage’ may limit the suitability of the gas since the fluorescent signal is easily saturated. The low diffusivity (0.07cm/s) of gaseous iodine explains its utility for flow tracking, but the difficulties of seeding at a constant rate (with levels typically at ~400ppm [288]) can hinder its experimental applicability [261]. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 46 3▪10▪3 ORGANIC MOLECULES The field of organic molecular photophysics is itself almost boundless. This is reflected in the vast body of work presented by Birks in his two volume Organic Molecular Photophysics [291, 292], and his reliance on the contributions of two dozen contributing authors and innumerable cited works. Polyatomic molecules exhibit wide broadband absorption since the large number of quantum interactions permits many ‘allowed’ states and transitions. The large number of excited states provides a correspondingly wide emission spectra. Matching the tracer to the fuel, primarily in terms of boiling point and mass diffusivity, should ensure (as much as is possible) that the mixing processes are similar in the two substances. 3▪10▪3▪1 AROMATIC COMPOUNDS Aromatic compounds are far less widely used than aliphatic ones (cf. §3▪10▪3▪2). The most widely used aromatic compound is toluene (C6H5CH3). Toluene absorbs broadband UV from ~220-280nm, peaking at 262nm; with emission over the range 265-350nm and a maximum intensity at 284nm [293]. Initial work by Reboux et al. [294] at 248nm excitation found a linear dependence of fluorescent intensity upon incident laser energy (up to 5mJ/mm2). The same study also found 5%v/v toluene in iso-octane to be optically thin, and not self-quenching. The effects of nitrogen, water vapour and CO2 upon fluorescent intensity were found by the same authors to be negligible. Oxygen was hypothesised to be toluene’s main quench partner (which was later studied in detail by Koban et al. [295]) and the direct relationship observed between intensity and oxygen concentration suggested a linear FAR-LIF relationship. However, Reboux et al. noted a temperature dependence of ~20%/1000C which was later used by the same researchers [294] to investigate charge inhomogeneity in SI-ICE’s. The purported results are quoted as being accurate to within 2%, yet the work of Koban et al. [296, 297], in the author’s opinion, casts doubt upon this assertion, in that the toluene signal does not scale linearly with oxygen partial pressure. Koban et al. [295] propose a detailed two step fluorescent model which accurately predicts the behaviour of toluene emission in the presence of oxygen at bath compositions relevant to ICE’s. Indeed, in 2002 Frieden et al. [298] used toluene’s oxygen dependent behaviour to predict oxygen concentrations in a directly injected spark ignited engine. For the current study however, toluene is of limited use since its physical properties lend it to tracking iso-octane (rather than methane), though the work of Fujikawa et al. [250] compares the use of toluene to acetone at 1bar and 200C and notes that at 248nm toluene is 90 times more fluorescent than acetone.  ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 47 Polycyclic aromatic hydrocarbons (PAHs) may be seeded into a flow or, as is more often the case, be part of the (commercial) fuel11 to be studied. Their absorption in the blue-UV range is complex, though well documented and noted to shift hypsochromically with increased pressure [292], Ch.3. Similarly, the emission, which occurs over a similar wavelength to the absorption (the details of which are heavily dependent upon the exact structure of the compound), is broadband and highly structured [292], Ch2. The large absorption cross section of PAHs and their high quantum yields (Φf = 0.17 – 0.82 [261]) makes them ideal tracers from a spectroscopic point of view ([299], [292] Ch. 2-3). The fluorescent lifetime of the emission may even be used to map the composition of the PAH ensemble, given that each is unique [300] and may be coupled with detailed knowledge of the hypsochromic shifts observed for increased compound size [261]. The large number of PAHs available allows tailoring of tracers for a bespoke fit to complex fuels; though ensuring a matched boiling point does not ensure fuel following behaviour [301, 302]. Further, any PAHs larger than toluene or benzene need to be seeded at above room temperature because of their low vapour pressure [261]. Thijissen et al. [192] have conducted in situ LIF PAH concentration measurements in an industrial scale fuel rich natural gas flame. Excitation using an argon ion laser at 488nm established a correlation between the LIF signal intensity and the PAH concentration. It was found that LIF emission is strongest from high molecular weight PAH’s (e.g. coronene), and that their fluorescence emission spectra are similar to those from flames. The difficulty of seeding PAHs directly and matching their mass diffusion properties to gaseous fuels has not been overcome, while the health concerns of these compounds also often precludes their use. 3▪10▪3▪2 ALIPHATIC COMPOUNDS Aliphatic compounds are easily excited in the UV though tend to photo-dissociate. Longer unsaturated or conjugated systems are generally unstable and are prone to polymerization [261]. The need for an easily accessible chromophore leads to structures like ketones (R2CO), aldehydes (R-CHO) or amines (R3N) 12, while conjugation is also know to produce marked hypsochromic shifts in the fluorescence spectra [261].  Ketones have been studied extensively for their spectral and photochemical properties. Their pressure and temperature dependencies are relatively easily modelled ([303, 304] and [305] respectively), and the fluorescent lifetime of ketones is well established [306] for a range of temperatures and pressures [307]. Knowledge of the role of oxygen in ketonic systems is also well established [308]. The combined effect of this extended understanding has allowed several authors to present detailed ketone photophysics [303, 309]. The range of accessible vapour pressures allows ketones to track gas phase fuels well (e.g. the use of acetone to track natural gas: [310, 311]); as well as liquid phase fuels where 3-pentenone (or a mixture                                                        11 Gasoline contains high concentrations of single ring aromatics (toluene and xylene) while diesel has plenty of two-ring compounds (naphthalene and its derivatives) 12 Where R is a saturated hydrocarbon ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 48 of 3-pentanone and 3-hexanone [312]) has been used widely to track gasoline [307, 313, 314], or more commonly iso-octane. Biacetyl (CH3(CO)2CH3) is widely used in fluorescent and phosphorescent studies since its boiling point of 880C matches well that of liquid fuels and in some cases can be deemed suitable for gaseous flows and it exhibits a high phosphorescent yield (~15% [261]) in oxygen-free flows. However, biacetyl has a low vapour pressure, which makes it difficult to seed.  Formaldehyde (CH2O) is difficult to dope since it polymerizes easily [261], though its temperature dependent fluorescent behaviour also allows the identification of temperature variations in the immediate pre-combustion field [190, 315-317]. Formaldehyde LIF is often used simultaneously with CH LIF for flame front imaging [318, 319]. Further, coupling of the CH2O and OH LIF signals has been shown by Paul et al. [237] to correlate well with the local heat release rate. The applications of CH2O LIF are widespread and accurate [190, 200, 243, 316, 317, 320, 321], but are tied to the combustion process as the method of CH2O production, and as such leave CH2O of no use in mixture formation diagnostics.  The next homologous molecule to formaldehyde is acetaldehyde (CH3CHO), which has been successfully used as a tracer by Arnold et al. [322]; however, acetone is commonly used in preference since it is less harmful. Aldehydes exhibit small quantum yields compared to aromatic compounds in the absence of oxygen, yet are comparable in the presence of air. Hexafluoracetone is suggested as an alternative to acetone since its quantum yield is an order-of-magnitude greater [323], though there are problems with its toxicity. Hansen and Lee [324] have investigated the radiative and nonradiative lifetimes of linear aldehydes. Radiative emissions share much the same lifetime as large aldehydes and follow well the Stickler-Berg equation, expect for acetlylealdehyde which has a much longer radiative lifetime than expected.   CH LIF provides similar information and is applied in much the same was as CH2O LIF. It is particularly useful in methane and natural gas combustion studies. It is, however, applicable across a wide range of hydrocarbon fuels due to its prevalence in hydrocarbon combustion pathways. Yet, without a flame the practicalities of generating and seeding CH prohibit its use in mixture studies when easier and more effective tracers are available.  Carbon dioxide LIF can be spectrally resolved in the 215-255nm range using an Nd:YAG pumped dye laser and has been successfully used by Lee et al. [325] to study CH4/O2/Ar and CH4/air flat-flames at fuel/air ratios from 0.8-1.9 in the pressure range 5-40bar. Emission is broadband in the range 200-450nm and has a weak structure. Further, the signal was found to scale linearly with laser fluence and pressure (within the range stated). However, in general, the signal from CO2 is weak and strongly temperature dependent, precluding its use as a direct measure of concentration [261]. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 49  Amines, as is common to most aliphatic compounds, fluoresce under UV excitation. They are however, effectively quenched by oxygen and often deemed prohibitively toxic for widespread use. Amines used in their own right include ethylamine [326] and N,N-dimethylaniline (DMA) [327], though they are more commonly used in exciplex systems particularly as aniline groups (i.e. NH2 on a benzene ring) [261].  3▪11 ACETONE AS A FLUORESCENT MEDIUM Acetone (also called propanone, dimethyl ketone, 2-propanone, propan-2-one and β-ketopropane) is an aliphatic ketone with formula (CH3)2CO, cf. Figure 3-3, and molecular weight 58.08. Its high vapour pressure (24 kPa at 200C) facilitates high seeding densities, while its vapour pressure behaviour fits the Antonie equation well [196]. The photochemical properties of acetone have been studied extensively, starting with the work of Heicklen [328, 329]  and Groh [330]. Typical of ketones, acetone exhibits strong absorption from 225-320nm, with a peak at 277nm [331], though the absorption spectrum is ostensibly flat between 270-280nm [196]. The emission spectrum of acetone is broadband blue between 350-550nm with peaks at 445 and 480nm [196]; the spectra is, however, affected markedly by vibrational relaxation prior to emission (cf. §3.11.4). Acetone has flammability limits which are easily accommodated and it is not carcinogenic or excessively toxic, though prolonged exposure is advised against.  Figure 3-3 Molecular Acetone, (CH3)2CO  Upon excitation, various process are competing to deactivate acetone, cf. Figure 3-4. From the first excited singlet state, S1, there may be: fluorescence, the spontaneous emission of radiation resulting from a transition which maintains spin multiplicity, S0←S1(n,π*);  intersystem crossing, the conversion from one state to another of different multiplicity T1(n,π*)←S1(n,π*) [309]; internal conversion, the non- radiative conversion from one state to another of equal spin, S0←S1(n,π*); or photodissociation of the acetone to several possible photo-products.  Figure 3-4 Deactivation Pathways for the First Excited Singlet, S1, of Acetone,  adapted from [332]      O C H3C CH3 ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 50 Acetone’s status as the de facto tracer for methane comes from its similarity to methane in the most salient of physical properties, cf. Table 3-2.      Methane  Acetone Molecular Weight (g/mol) 16.043 58.080 Density @ 250C (g/cm3) -- 0.786 Boiling Pt. (0C) -161.48 56.1 Lower Heating Value (MJ/kg) 50 28.6 Heat of Vapourisation @ 250C (kJ/mol) -- 30.99 Heat of Combustion (MJ/mol) 0.8908 1.8207 Max. Burning Velocity  @ 250C with Φ Specified (cms-1) 44.8 @ 1.08 44.4 @ 0.93 Flash point (0C) -188 -18 Autoignition Temp. in Air (0C) 537-632 465-727 Flammability Limits in 1 bar Air (% vol) 5.0-15 2.6-13 Gas Phase Viscosity at 1000C (µPa s) 13.4 9.5 Gas-Phase Diffusion Coefficient in 1 bar air @ 1000C (cm2s-1) 0.344 0.166 Gas-Phase Diffusion Coefficient in 8 bar air @ 1300C (cm2s-1) 0.0493 0.0239 Table 3-2 Physical & Thermodynamic Properties of Methane and Acetone Gas, excerpts from [259, 261] and references therein. 3▪11▪1 EXCITATION SOURCE CHOICE The broad absorption spectrum of acetone facilitates the use of many excitation sources. It is important to note that as the absorption cross-section increases for shorter wavelengths, the fluorescent efficiency decreases slightly due to higher dissociation (cf. §3▪11▪3) The ‘figure of merit’ (sic, [196]), Eσ, in the final column of Table 3-3 is the product of pulse energy and absorption cross-section at the laser wavelength, and provides a metric against which to measure the efficiency of each excitation source. With the exception of the difficult to use flashlamp pumped rhodamine 590 dye laser the krypton fluorine (KrF) laser shows the best applicability, and as such is used in the current work.  Laser λ (nm) Energy/pulse (mJ) Eσ (x10-20 Jcm2) XeCl excimer 308 300 0.48 KrF excimer 248 300 0.7 Quadrupled Nd:YAG 266 120 0.52 Flashlamp Pumped Rhodamine 590 280 400 1.9 Raman Shifted ArF Excimer 254 2 0.006  284 1 0.0045 Table 3-3 Acetone Excitation Sources,  after [196]    ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 51 3▪11▪2 ABSORPTION Upon excitation the singlet ground state of acetone is excited to the first electronic state of similar spin multiplicity [333], which is 30435 cm-1 above the ground state [334] (although Thurber et al. [305] use 30440cm-1 in their modelling work). The chromophoric C=O bond undergoes a transition to elevate an electron from its non-bonding, n, oxygen orbital to that of an anti-bonding π* orbital, leaving the electron unpaired and the oxygen with a partially filled p-type orbital (cf. Figure 3-6). The n→π* transition is orbital-forbidden though spin-allowed [307], and is thus weak. In its excited state the weakened C=O bond leads to carboxyl out-of-plane wagging and stretch [333, 335], while also causing molecular lengthening and pyramindization at the carbonyl carbon atom through increased C-C torsion [309, 333, 335]. The low energy of all these behaviours means that their frequencies are much lower than that of the molecular vibration, obscuring any rotational structure in the absorption band for acetone [336, 337], cf. Figure 3-5. A hypsochromic shift, of about 2nm/100K, in the absorption spectra for 3-pentanone (which, as (C2H5)2CO is a corollary of acetone) is noted by Grossman et al. [307]. These findings were later extended and confirmed to apply to acetone by Thurber et al. [305, 306] and Ossler & Aldén [306], the former also noting that the absorption cross-section increase significantly with temperature. The increased absorption cross-section roughly compensates for the loss in cross-section caused by the hypsochromic shift for excitation wavelengths shorter than the peak at 277nm (at ambient temperatures).  Figure 3-5 Absorption Spectrum of Acetone at Room Temperature and 1 atm, reproduced from [196]    3▪11▪3 PHOTODISSOCIATION OF MOLECULAR ACETONE The cleaving of the α-CC bond occurs since the bond overlaps the vacant non-bonding oxygen orbital in the excited π* state (Figure 3-6) [338]. The latest ab initio studies by Liu et al. [339] suggest that the dissociation barrier for the lower energy methyl acetyl split, Fig.3-6a, lies at 556 kJ/mol above the ground sate. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 52  Figure 3-6 Acetone Dissociation through α-Cleavage,   adapted from [338]     This cleavage reaction allows for multiple products, though the methyl acetyl split is more thermodynamically preferable [338], and explains the predominance of Eq.3.1 in the photodissociation of acetone [340, 341].   Eq. 3.1 For excitation at less than 312nm [196], and particularly under 300nm [342], the alternative Eq. 3.2 is more likely.  Eq. 3.2 It is noted by Liu et al. [339] that the energies required to form the products of Eq.3.2 are unattainable by direct 248nm excitation, and a step wise decomposition of the (CH3)2CO molecule is more likely [343]. Consequently, at wavelengths below 250nm the reaction outlined by Calvert & Pitts [344] and generalised by Gilbert [338] may also occur (Figure 3-6(b)), viz.:  Eq. 3.3 Haas [309] proposes a detailed 11 step mechanism for the photodissociation of acetone, which outlines exact product formations, including that of biacetyl (cf. §3▪11▪3▪1). Observed reductions in phosphorescent intensity with decreased wavelength indicate that photodissociation increases with excitation energy [345]. Excitation of acetone at 248nm through the global step of Eq. 3.1 leaves many (~30%) [343] of the photofragments with enough excess energy that the acetyl radical may spontaneously decompose to form, with the original methyl radical, the products of Eq.3.2. The linear Stern-Volmer plot obtained from the data of Blitz et al. [342] for λ<300nm suggests deactivation from only one excited state- the singlet state S1- according to the scheme (that may include any third body, M):  Eq. 3.4 CH3 H3C C O • • CH3 H3C C O • • H3C C O • • + CH3 hv (n,π*) CH2 H3C C O • + • H CH3 H3C C O • • CH3 H3C C O • • hv (n,π*) (a) (b) ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 53 3▪11▪3▪1 BIACETYL POISONING The main products of acetone photochemical pyrolysis are: ethane, methane and carbon monoxide [309]; however, the acetyl radical resulting from Eq. 3.1 is important since it may recombine to form biacetyl, (CH3CO)2, which produces a strong phosphorescent signal [346]. This competes with the fluorescence of the acetone S1 state, and may complicate the collection of a high fidelity fluorescence signal. Biacetyl formation from triplet acetone is only of concern for excitation above 300nm [342] or at temperatures below 1000C [347]. However, when present biacetyl effectively quenches the T1 state of acetone, reducing the tendency for acetone to phosphoresce, but in the process exciting itself to phosphoresce (at wavelengths similar to acetone) [328]. Although biacetyl triplets are known to phosphoresce with high efficiency (~15%) they are also effectively quenched by oxygen; while biacetyl fluorescence has a very low efficiency below 320nm [196]. For excitation at 248nm it is observed that there is a move away from Eq.3.1 towards that of Eqs. 3.2 & 3.3 reducing the biacetyl concentration to a level which is not of concern [309]. While, under thermal conditions the low concentration of acetyl radicals negates the need to worry about biacetyl formation [309]. All of these features mean that for the oxygenated environment of the RCM, biacetyl poisoning may be disregarded. 3▪11▪3▪2 THERMAL DISSOCIATION Thermal pyrolysis of acetone, through the Rice-Herzfield mechanism [348], primarily yields carbon monoxide, methane and ketene ([349] in [309]), with small amounts of butanone, ethane and 2,5- hexanedione [309]. However, the process is not significant for temperatures below ~800K, which the compression stroke of the RCM fails to generate. 3▪11▪4 NON-RADIATIVE DEACTIVATION  Throughout the following discussions reference to Figure 3-7 may be of help to the reader. Hansen and Lee [332] assert that “…the non-radiative processes predominate over the radiative processes at least by two orders-of-magnitude in simple ketones.”, the low zero-pressure fluorescence quantum yield of ~0.002 reported by Heicklen [328] and Shortridge et al. [350] supporting such a view. Inter-system crossing from the singlet state is known to peak at intermediate energies [345], which suggests competing pathways for the deactivation of the S1 state based upon its vibrational state [303]. Rapid inter-system crossing (ISC) (kISC between 4 x 10 8 s-1 [350] and 3 x 108 s-1 [303], and [345]), depopulates the first excited singlet state to that of the first triplet manifold at  28000cm-1 above the ground state [334]. Although the process is spin-forbidden the first singlet and triplet states of acetone are closely coupled such that ISC efficiencies approach 100% [351-353]. The small energy difference between S1 and T1 means that collisions also effectively populate T1 from S1 [354]. The resultant mix of excited acetone in its singlet and triplet states has a predominant triplet character because of the efficient ISC [354]. The triplet ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 54 component may come from increased excess energy above the S1 state, which increases molecular vibration and rotation to promote ISC to the T1 state, or by direct optical preparation [354]. The n→π* transition is known to produce a movement of electron density away from the oxygen molecule, with the methyl moieties able to inductively donate electron desnity, stabilising the excited state [338]. This allows ISC to take place (potentially following vibrational relaxation, VR, in the excited singlet state) before rapid, fluorescent deactivation that is independent of the bath phase [303].  The vibrationally hot triplet state, T1 **(which predominates in the high pressure and liquid phases of acetone [354]) may phosphoresce (τT1** ~ 30µs [345]), dissociate or relax to a thermally equilibrated state, T1 [355], where it cannot be fully quenched due to its high reactivity [309], though may still dissociate through the Rice-Hertzfeld mechanism [348] (cf. §3▪11▪3▪2), or phosphoresce. Bitto [356] asserts that it is the vibrational coupling of the torsional modes in the symmetrically favourable triplet manifold which increases the lifetime of the triplet state. This thermal triplet, T1, does not however correlate directly with the triplet ground state [309] and so no fluorescence is observed from a possible T0←T1 transition.   Figure 3-7 Jablonski Diagram for a Electronically Excited Organic Molecule,  reproduced from [261]. 3▪11▪5 RADIATIVE DECAY The mixed singlet-triplet state leads to a quasi-biexponential decay of the excited state [337] whereby the nature of the emission is governed by the rovibronic interactions of the molecule [354]. When excited to near the S1 origin, decay to the S0 ground state is most likely through fluorescent emission [354]. Yet for stronger stimulation the small number of molecules left in the S1 excited state after ISC may relax to a vibrationally excited ground state, S0 *, through internal conversion (IC). Those molecules left in the metastable T1 state (populated directly from ISC or thermally relaxed T1 ** molecules) have a lifetime which is purported by Bitto [356] to be extended due to symmetrically favoured vibrational coupling. This accounts for the long lifetime observed for the phosphorescent relaxation to the ground state S0 (τT1 ~ ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 55 200µs [345] & 196). The emissions spectra of acetone are shown in Figure 3-8 where the lower energy of the triplet emission shows a clear hypsochromic shift.  Figure 3-8 Fluorescent (solid line) and Phosphorescent (broken line) Emission Spectra for Pure Acetone, reproduced from [331]    3▪11▪6 FLUORESCENT SIGNAL LINEARITY, QUENCHING & THERMODYNAMIC STATE EFFECTS For weak excitation of acetone, the fluorescent intensity is known to be proportional to incident laser intensity (up to ~1 J/cm2) [196, 357]. To remain linear however Lozano et al. [196] assert that the inequality Eq.3.5 must hold true.  Eq. 3.5 For 248nm excitation the absorption cross-section and lifetime yield Ni to be ~ 4x10 28 photons/cm2s. For a 20ns pulse at this wavelength, incident laser intensities must reach ~630 J/cm2 to saturate the signal, yet the KrF excimer used in the current work can only supply a maximum of 166 J/cm2, implying that all fluorescent intensity values may be deemed proportional to laser beam energy. Copeland and Crosley [336] assert that the lowest photodissociation barrier for acetone is at 360kJ/mol above the ground state. For a constant 1mm thick laser sheet in the RCM cylinder, with an acetone concentration of 1%, the maximum laser power (of 400mJ) would excite the acetone to ~42kJ/mol above its ground state. This suggests that 248nm excitation avoids the need to consider dissociative effects- an assertion supported by the work of Liu et al. [339]. Finally, the fluorescent intensity observed is also known to be proportional to the partial pressure of acetone vapour [196, 357].   The failure to generate a Boltzman population of the probe volume in the S1 state upon excitation (because of the fast ISC from the singlet to triplet states) yields the prompt fluorescent emission observed, which in turn limits the susceptibility of singlet acetone to quenching [258, 358]. Oxygen, which has a triplet ground state, is the notable exception to this. The small but distinct quenching effects of oxygen upon acetone fluorescence are reported by several authors [304-307], and are well understood for excited molecules in general [330, 359, 360]. The triplet state of oxygen is a more effective quencher than its singlet equivalent [308], which may result from its energetic similarities to the 1(n,π*) state of acetone. ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 56 The consequence is that oxygen quench rates are close to, or exceed, the gas kinetic rate constant for acetone [354]. For ketone oxygen interactions Grossman et al. suggest the mechanism of oxygen quenching to be based upon “…the formation of a charge transfer complex with the ketone, involving the electronegative oxygen atom of the excited carbonyl bond and the electrons of the [oxygen] double bond.” [307]. The work of Nau et al. in the liquid phase suggests a similar mechanism to that of Grossmann et al. with the formation, upon ketone-oxygen collision, of an encounter complex which if stable will give an overall quench rate proportional to collision frequency ([361] in [357]). Breuer et al. [303] suggest the findings of Nau et al. to be due to oxygen enhanced ISC that serves to depopulate the S1 state. They correspondingly adjust the work of Thurber et al. [305] to correct for fluorescence yield at shorter (248nm) wavelengths, though no significant new trends were observed. Excitation of acetone at shorter wavelengths is noted to increase the likelihood of quenching since further excitation above the S1 origin allows for more vibrational relaxation before ISC or radiative deactivation [307]. The vibrational level of the excited state is also known to have a small effect on the encounter probability of oxygen assisted ISC [357].  In the gas or vapour phase the effects of quenching manifest themselves through differences in fluorescent intensity at different temperatures and pressures. The work of Thurber & Hanson [357] indicates that in a nitrogen bath gas fluorescent intensity increases asymptotically towards a high pressure limit (at ~16.5atm for 248nm excitation)- attributable to full vibrational relaxation of the S1 state prior to ISC. This limit will be reached sooner at longer excitation wavelengths since there is less vibrational relaxation to take place before radiative decay. Similar trends are observed in a pure oxygen bath gas, with the high pressure limit being reached sooner (at ~4atm) because of more effective oxygen quenching (though the fluorescent intensity is also observed to decrease for increases in pressure past the upper limit). The data of Grossmann et al. [307] which pertains to a synthetic air bath gas, unsurprisingly, shows a tendency to reach a high pressure limit (at ~8-10 bar) which sits between the two cases of Thurber & Hanson. The behaviour is not paralleled exactly however. For 248nm excitation an increase in fluorescent intensity is observed up to eight bar, where the high pressure limit is apparently reached, yet past ~25 bar a steady drop of ~1% per bar is then observed without explanation (up to an experimental maximum of 50 bar). These findings are at odds with those of Yuen et al. [304] who, for their air bath studies, report the drop in fluorescence to start at ~5 bar and continue until the experimental maximum of 8 bar.   Temperature affects emission by varying the product of absorption cross-section and fluorescent yield for a given molecule. Koch and Hanson report no significant change in intensity for 3-pentanone fluorescence in air over a range of bath temperatures [358]. This is consistent with the observation, in another study by Koch & Hanson [362], that acetone has an almost constant absorption cross-section at ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 57 248nm excitation (the absorption cross section was observed to rise by only ~5% over a 600K temperature increase). However, other significant works [305, 307, 363, 364] show small though clear dependencies of acetone and 3-pentanone emission on temperature. Grossman et al. [307] report (for 248nm excitation in synthetic air) a steady drop in fluorescent intensity of ~35% per 100K while Ghandi & Felton [363] report similar, though slightly less dramatic, findings for acetone in a heated jet and a motored engine (at 266nm). All the studies are able to discount thermal dissociation for the decrease. The findings of Thurber et al. [305] are in agreement with those of Tait & Greenhalgh [364] (who use 308nm excitation of acetyhldehyde) and Grossman et al. [307], and extend the results from a 600K experimental maximum to 975K. Beyond ~800K the drop in fluorescence is seen to reduce by ~8% per 100K, with a smooth transition between the two regimes. Although there is a significant increase in absorption cross-section with temperature it does not correct for the decrease in fluorescence and so a marked drop in fluorescent yield is observed for increased temperatures [305], and as such is a useful temperature diagnostic.  According to the model of Thurber & Hanson [357], when the above tendencies for pressure and temperature are combined for acetone, at 248nm excitation, a rise in fluorescent yield is still observed. Indeed, it is noted that pressure effects are stronger for the shorter wavelengths and the reduction in fluorescent yield due to temperature rise mitigate only slightly the rise due to pressure. For increased pressure and temperature Ossler & Aldén [306] report similar effects on fluorescence decay time to those of fluorescence intensity. From a ‘zero-pressure’ minimum up to ~10bar the fluorescent lifetime increases linearly from ~0.9ns (with a slope of ~0.15ns/bar at 323K decreasing to ~0.02ns/bar at 723K) as in Figure 3-9.  Figure 3-9 Fluorescent Lifetime Variation of Acetone with Pressure, reporoduced from [306]  3▪12 CONCLUSIONS The most salient details of the main techniques that could be used to study species concentration have been outlined. The suitability of each to combustion diagnostics has been discussed, and from this the utility of laser induced fluorescence made patent. The versatility and diversity of LIF applications has ▪ Chapter 3 ▪ Laser Imaging Techniques & PLIF Fundamentals ▪ 58 been outlined and a comprehensive discussion of possible gaseous tracer compounds made. A strong case for the use of acetone as a proxy for methane/natural gas has been made, primarily on the basis of their well matched physical properties. Further discussion has been offered as to the utility of acetone, its spectral properties, and how these properties vary with engine relevant conditions. Acetone LIF has been studied extensively and its behaviour is well understood, so allowing the contemporary experimentalist to draw meaningful and accurate inferences from its use.    59 Chapter 4   EXPERIMENTAL SETUP & PROCEDURES    4▪1 INTRODUCTION Chapter 4 offers the most important details of the rapid compression machine used in the current study. Important changes are highlighted, primarily with respect to the improvements that allow combustion studies to be supported and the control systems required to operate the fuelling and laser diagnostics systems. The experimental test matrix is outlined, with all pertinent parameters and operational details provided to the reader, along with the rational behind the choices made.  ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 60 4▪2 THE RAPID COMPRESSION MACHINE The rapid compression machine, RCM, was designed and built by Döhring [365] and can simulate any two consecutive strokes of an internal combustion engine. The RCM is shown schematically in Figure 4-1. The most important features are considered in detail below.  Figure 4-1 Schematic of Rapid Compression Machine (RCM)  The principle of operation is as follows. The RCM piston and con-rod are driven by a vertical crank shaft, which is itself driven by a rack, where the rack is actuated by a pneumatic cylinder. The braking force is provided by a hydraulic piston on the opposite side of the rack. The pneumatic cylinder is supplied by an 80gal accumulator, which is charged from the regulated laboratory compressed air supply. The braking side is a closed loop system filled with 20W30 oil. When air pressure is supplied to the driver cylinder, the rack is prevented from moving by the solenoid valve on the braking side. This valve prevents flow in the oil loop. To actuate the RCM the solenoid valve is triggered so that the oil in the braking cylinder is allowed to flow up into the reservoir. When the rack is returned to the start position, a small secondary loop is opened on the braking side such that the braking piston is allowed to fill from the reservoir. The details of RCM operation have been covered previously [365, 366], and only those operational changes pertinent to this study shall be reported herein.  The single shot nature of the RCM, its lack of intake and exhaust manifolds, and the omission of the valve train mean that cyclic variability is almost eradicated, whilst also ensuring strict control of the in-cylinder Braking Section Cylinder Driver Section ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 61 charge motion. Any motion may be fully eliminated to give a quiescent starting state from which all subsequent fluid movement is imparted by the piston. 4▪3 NEW RCM CYLINDER  As part of the current work the rapid compression machine was altered significantly from its previous incarnations. Prior studies using the RCM [365, 366] have been based on cold flow. The PMMA cylinder arrangement of the RCM was redesigned and retrofitted with a UV grade HPFS (high purity fused silica) cylinder and fire deck, and a new 316 stainless steel cylinder and cylinder head installed to allow combustion (cf. Figure 4-2).  Figure 4-2 Exploded View of New RCM Cylinder 4▪4 RCM FUELLING SYSTEM The RCM is set up to have its mixture prepared from three different sources. A premixer is used to prepare homogenous NG/air mixtures based upon partial pressures. The premixer is instrumented for NG line pressure, total pressure and mixture temperature. This premixed charge can be supplied directly to the RCM cylinder for homogenous fuelling, or may also be supplied to the PSC capillary for premixed PSC ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 62 studies. The PSC capillary tube, which has its injection controlled by an Omega SV-121 high-pressure solenoid valve, can also inject pure methane. The final source of fuel for the RCM cylinder is the J43M direct injector, which can only inject methane into the RCM cylinder.  The laser induced fluorescence technique used in this study requires that acetone be mixed with NG fuel. The high-pressure seeder used to achieve this is covered in §4▪5 The seeded NG may be sent to the PSC the direct injector or to the premixer. This allows LIF studies to be carried out on any combination, of the fuel sources.  The piping diagram to achieve the abilities outlined above is shown in Figure 4-3, while its details are presented in Appendix A.   Figure 4-3 Schematic of RCM Fuelling System  The mixture within the RCM cylinder can be ignited with a spark between tungsten electrodes. The electrodes have a 40 angle w.r.t. the fire deck to clear the PSC capillary tube. The electrode position also has a small amount of adjustability so that the location of the spark within the PSC plume or homogenous charge can be adjusted.  N 2 C H 4 A IR   VENT T T S E E D E R   P R E M IX E R   T Thermocouple Pressure Gauge Check Valve Valve  Regulator Relief Valve  2W- 2 Way 3W- 3 Way 5W- 5 Way B- Ball N- Needle 3W-B N N N VENT SAMPLE   J43 VENT VENT SAMPLE 3W-B 3W-B 5W-B 5W-B 5W-B 3W-B  3W-B   PSC  2W-B  2W-B   2W-B RCM  2W-B N N N 2W-B 2W-B VENT ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 63 4▪4▪1 THE WESTPORT® J43M INJECTOR Much of the information pertaining to the J43M natural gas injector is proprietary and confidential. The injector was developed by Westport Innovations Inc. as part of its own R&D program and was not a commercial product. The actuation of the injector needle is achieved through the energizing of a magnetorestricitive material. Upon imposition of a magnetic field to the Terfenol-D™, the magnetic regions within this ferrous material realign which causes the volumetric expansion which actuates the needle. The movement of the needle allows the gas supply (at up to 3600psi) to flow through the injector. The injector is PWM controlled, however the injector requires a 5V logic signal for the entire injection duration. Further, rate shaping is possible by varying the logic level of the needle lift signal. In the current work no rate shaping was required and maximum needle lift was always specified.   Two J43M injectors were originally specified for the current study: one to supply the direct injection fuel plume and one the partially-stratified-charge injection. The characterisation information for both injectors is listed in Chapter 5. However, persistent technical problems with the PSC injector necessitated its replacement with a high-pressure (1000psi) solenoid valve (Omega SV-121). The solenoid was connected directly to the capillary tube supplying the RCM cylinder. The implications of this change for the current study are minimal in that it is only the initial stages of the injection that are relevant to PSC DI interaction. However, the overall reliability and repeatability of the response of this model of solenoid have posed significant problems to other researchers [7, 367].  Provision of the PSC charge can also be made through the use of an ‘insert’ that replaces the PSC capillary tube. The RCM cylinder quartz window in the cylinder head may be replaced with an aluminium equivalent which is threaded with a ¾” parallel thread. This thread accepts the insert, which hold a 14mm sparkplug. The insert, pictured in Figure 4-4, has natural gas supplied by the same solenoid as for the capillary tube; however, the insert now provides a flow path to direct the fuel in front of the plug electrodes. ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 64  Figure 4-4 PSC Insert 4▪5 FLOW SEEDING As was alluded to previously, the LIF studies of the current work require the fuel to be doped since natural gas does not naturally fluoresce. This is achieved with the high-pressure bubbler shown in Figure 4-5, which is based on the design presented by Neij [259], and adapted for higher pressures. The detailed design information for the seeder is presented in Appendix A, though the basic principles of operation are outlined here. The vessel is a thick-walled 316 stainless steel chamber with flanged ends. There are a number of tapings to the vessel: two gas inlets, one gas outlet, a refill inlet, a pressure relief valve, pressure gauge and a thermocouple tapping (shielded type-K). The last two provide state information for the seeder, while the former allows methane to be fed into the seeder in batches (based on supply pressure) or continuously. The two inlets vary the flow path. One inlet bubbles gas through the liquid acetone, while the second inlet (not shown in Figure 4-5) bypasses the liquid acetone. The two flow paths then mix as they pass through a bed of glass beads before effluxing. The flowpath of the feed gas is shown in blue in Figure 4-5. This arrangement allows the acetone concentration in the effluent to be varied according to the fraction of the flow which was bubbled through the acetone.   ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 65    Figure 4-5 Section View of RCM Seeder 4▪6 LASER & OPTICAL SYSTEM A UV laser system is used to excite the acetone doped gas. 248nm UV light, with a spectral width of 3nm at full-width half maximum is supplied by a CompexPro 102 krypton-fluorine (KrF) excimer laser manufactured by Lambda Physik (now Coherent Inc.). The beam dimensions are 12mm x 24mm at the exit aperture, with a divergence half-angel of 3mrad. The beam path to the RCM and its formation into a light sheet are shown in Figure 4-6.  The beam is immediately sampled with a Melles-Griot 248nm beam sampler. About 16% of the beam energy is directed to the power meter (cf. §4▪7▪1, below), the remainder going to a broadband metallic mirror in a kinematic gimbal that turns the beam though 900 horizontally to a beam steerer. The beam steerer steps the beam up by approximately 300mm, with the beam exiting in line with the RCM cylinder axis. Two 50.8mm x 50.8mm UV grade cylindrical lenses (fl1=200mm & fl2=300mm) are then used to form a Galilean telescope. The light from the telescope is directed onto a high-power dielectric mirror (that is both wavelength and direction specific in its reflectivity) mounted on a kinematic gimbal. This mirror turns the light through ~98.50 so that the beam enters the cylinder at an angel of ~8.50. This angle allows the majority of the optical section of the cylinder to be illuminated, while also minimizing the area obscured by the beam impinging upon the PSC capillary. Adjustment of the final cylindrical lens and the high-power mirror allow the beam to set-up so that it is on the cylinder axis and vertical. By changing the length of the telescope it is possible to have the beam waist (the optically invariant minimum thickness) of ~0.8mm on the central axis of the cylinder. However, a lack of  Pressure Relief  Thermocouple   Glass Beads   Acetone ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 66 collimation after the 300mm lens means the beam is ~1.45mm thick at the cylinder bore. The beam is finally collected, to prevent the scattering of potentially dangerous UV light, by a beam dump under the RCM cylinder (not shown). LIF luminosity is collected with a Cooke Corporation DiCam Pro intensified CCD camera, fitted with a Nikkor AF Micro105mm, 1:2.8D, lens and B+W F-Pro UV filter.   Figure 4-6 Laser Sheet Formation 4▪7 EVENT & TIMING CONTROL The details of the RCM event and timing control system are offered in Appendix B. Here a brief overview is presented to provide some context for subsequent discussions. Figure 4-8 shows the timing diagram for the pertinent RCM events. Those events are: Event Pulse or Step PSC injection Step Direct injection Step Camera trigger Pulse Laser trigger Pulse Table 4-1 RCM Timed Events  All the events listed in Table 4-1 are timed from the crank angel encoder, which gives a single pulse every 0.036CAD. The RCM is triggered by actuating a solenoid valve on the braking side that allows the rack to move and which causes the crank angle encoder to rotate. The 5V TTL pulse train from the encoder is gated so that no pulses pass the gate until an index pulse is reached. For rotation after this index, pulses are allowed to propagate. This system is used to circumvent the variability in the rack start position. The pulse train is supplied to the RCM control box which counts the number of pulses in the train and compares it against the set start value for each event. When the value is reached, a positive TTL trigger pulse is ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 67 released by the control box for that event. This signal is a pulse, the duration of which is set internally by an RC constant (at ~10ns), or a top-hat with a set duration (in encoder counts).   Figure 4-7 RCM Event Timing Diagram 4▪7▪1 INSTRUMENTATION Further to the control system, the RCM is instrumented with a high dynamic response piezoelectric pressure gauge (PCB Electronics 0112A), which is connected to a charge amplifier (Kistler 5010) that has its output voltage displayed on an oscilloscope (Tektronix 2014B). The oscilloscope is connected to a PC with OpenChoice Desktop installed that allows the trace to be electronically recorded. The PC is also equipped with the camera control software (CamWare and CineControl). In the case of the LIF studies an iCCD (Cooke Corporation DiCam Pro) was used, while for the flame propagation investigation a high frame-rate CCD was used (Phantom). The final use for the PC in the RCM control system is to record and display the reading from the laser power meter (Coherent FieldMax Pro II) using proprietary FieldMax II PC software. ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 68  Figure 4-8 RCM Instrumentation Schematic 4▪8 TEST MATRIX 4▪8▪1 FUELLING RATIONAL The work of Gorby [7] on the use of PSC with late DI was marked by difficulty in getting the PSC plume to ignite reliably. The problem of poor inflammation was resolved if a weak background mixture was induced through a traditional port-fuelled approach. The minimum RAFR found to produce reliable PSC ignition was 1.79, thus the value forms the basis for the current study. Centering the central-composite designed text matrix (cf. §4▪8▪3) on this value, or close to it, allowed the investigation of the effect of the bulk charge fuel concentration. The current work used bulk RAFR’s close to the threshold value reported in [7], and to the rich and lean sides of it.  The fuelling levels for each of the three fuel sources in the RCM cylinder (i.e. the background bulk charge, the PSC injection and the direct injection) were chosen to provide an overall RAFR of one. The PSC injection was held constant at 3%w/w of the total charge. The DI injection duration was altered to maintain the λ=1 condition after accounting for the fuel in bulk charge. 4▪8▪2 TIMING RATIONAL The test matrix was designed to meet the objectives listed Chapter 1. The overarching aim of the study was to understand the interaction of the PSC fuel plume with that of DI fuel jet. As discussed in Chapter 2, the structure of gaseous jets is well-established and Turner’s vortex ball model predicts two distinct regions to any given starting jet. The head of the jet is comprised of a ball, or vortex, which grows in size through leading-edge entrainment and mass supply from the trailing steady-state jet. It is thus apparent Injector Driver Timing Control Trigger Gate PSC DI CCD CA Encoder Laser Pressure Trace Ignition Circuit ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 69 that the PSC fuel jet (which itself must have the same structure, although on a smaller scale) can interact with the DI jet in three distinctly different regions. The PSC jet may intersect the DI fuel jet: i) along its leading edge, ii) at the centre of the vortex-ball iii) following the vortex-ball in the steady-state jet. Case ii is shown schematically in Figure 4-9 with the expected rich regions shown in red, the stoichiometric ocntour in yellow and the lean areas in blue.  Figure 4-9 Partially Stratified Charge Jet and Direct Injection Jet Interaction  To achieve this the PSC injection timing was held constant at 120CAD BTDC (duration 6CAD) so that the nature of the PSC plume would remain constant regardless of its subsequent interaction with the DI jet. Further, all images were taken at the same time of 90CAD BTDC. This ensured the cylinder conditions were, within experimental error, the same for each image case. It must be made clear that these timings were not optimised for potential combustion quality, but were made on the grounds of obtaining the most information to meet the objectives of the study. The image timing of 90CAD BTDC is advanced when compared to the late injection regime of a typical DISI engine. An understanding was sought for the full extent of the DI and PSC plumes, and use of a timing later than 90CAD BTDC would have had the piston impinge upon the image region. 4▪8▪3 CENTRAL COMPOSITE DESIGN Appendix E contains the mathematical background to the use of central composite designs and how they offer a statistically rigorous way of avoiding a full factorial test matrix. The two experimental variables (background charge AFR and DI injection timing), are mapped to the design variables x1 and x2, where - 1<x<1. Given that the design variables are at the limit of the experimental equipment range, the CCD must be inscribed. The inscribed CCD yields the data point locations of Figure 4-10, which correspond to the values given in Table 4-2. At each data location 51 repeats were taken, which, by the central limit theorem [368], allows a normal distribution in the responses to be assumed. Further, at the central run ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 70 location ((x1,x2) = [0,0]) three repeats were made (of 51 runs each) to minimise the response variance across the entire design space.  Data Point Location x1 x2 ξλ ξtime 1 -1 0 1.5 0 2 -1/√2 1/√2 1.57 3.52 3 0 1 1.75 5 4 1/√2 1/√2 1.92 -3.52 5 1 0 2 0 6 1/√2 -1/√2 1.92 3.52 7 0 -1 1.75 -5 8 -1/√2 -1/√2 1.57 -3.52 9 0 0 1.75 0 Table 4-2 Data Point Locations in Design (x) and Experimental Space (ξ)   Figure 4-10 Central Composite Design for Test Matrix 4▪8▪4 OPTICAL SET-UP PROCEDURE A set-up jig was manufactured that allowed a 0.5mm2 grid to be held on any longitudinal plane of the RCM cylinder. When horizontal it was used to provide a reference plane upon which to focus the laser sheet forming optics. The separation between the two lenses of the Galilean telescope was varied until the sheet -waste was on the RCM centre-line. Aligning the grid vertically allowed the camera focus to be set coincident with the laser sheet (any slight difference was accommodated by the 4mm depth of field), and the spatial calibration of the camera sensor. ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 71 4▪8▪5 EVENT TIMING PROCEDURE Initially, a Berkley Nucleonics 565 pulse generator was used to establish rough estimates for the time behaviours of the PSC and SI injection events. Under atmospheric conditions (without firing the RCM) LIF images were captured to asses the delay between commanded injection and the emergence of a fuel jet from the PSC capillary, and the extent of DI jet penetration at different image times.  Using the time estimates from the pulse generator study, a series of initial set-up runs were performed to match the jet interactions according to the design matrix, ultimately establishing the (x1,x2),=[0,0] case and the (x1,x2) =[0,1] case; the other points following from the constraints of a rotatable CCD or experimental bounds. The results of the timing set-up procedure, shown in Table 4-3 to Table 4-5, form the data points of the CCD as previously described, with injection durations varied to maintain an overall  λ=1 (Table 4-5) Start Duration Event CAD BTDC Encoder Counts CAD BTDC Encoder Counts PSC injection 138 1411 6 167 Direct injection variable variable variable variable Camera trigger 90 2744 --- --- Laser trigger 89.8 2850 --- --- Table 4-3 Control Box Capillary Injected PSC Event Timings  Start Duration Event CAD BTDC Encoder Counts CAD BTDC Encoder Counts PSC injection 90 78 6 167 Direct injection variable variable variable variable Camera trigger 90 2744 --- --- Laser trigger 89.8 2850 --- --- Table 4-4 Control Box Insert Injected PSC Event Timings  DI Injection Start Time ξtime DI Injection Duration Data Point Location ξλ CAD  (BTDC) Encoder Counts CAD  Encoder Counts 1 1.5 117.36 1984 4.152 115 2 1.57 113.83 2082 4.642 129 3 1.75 112.36 2123 5.640 157 4 1.93 113.83 2082 6.437 179 5 2 117.36 1984 6.629 184 6 1.93 120.89 1886 6.437 179 7 1.75 122.5 1842 5.640 157 8 1.57 120.89 1886 4.642 129 9 1.75 117.36 1984 5.640 157 Table 4-5 Data Point Experimental Setup Values  ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 72   Background charge compositions were mixed according to the partial (gauge) pressures of Table 4-6 Total Pressure (psig) NG Partial Pressure (psi) λ =1.5 λ =1.57 λ =1.75 λ =1.92 λ =2 30 445 465 515 565 585 35 520 545 605 660 685 40 595 625 690 755 785 45 670 705 780 855 885 50 750 785 865 950 985 55 825 865 955 1045 1085 60 900 945 1045 1140 1185 65 975 1020 1130 1240 1285 70 1055 1100 1220 1335 1385 75 1130 1180 1310 1430 1485 80 1205 1260 1395 --- --- 85 1285 1340 1485 --- --- 90 1360 1420 --- --- --- 95 1435 --- --- --- --- Table 4-6 Homogeneous Charge Partial Pressure (Gauge) Values Assuming Prior Air Flush. 4▪8▪6 MISCELLANEOUS EQUIPMENT SETTINGS Equipment  KrF Laser  Mode: Constant Energy, 300mJ Trigger: External Gas Supply: Partial Gas Replacement DiCam Pro  MCP Decay: 2ms MCP Gain: 99.5% Analogue Gain: Normal Trigger: External ROI: 17:29 × 5:29 Exposure: 1200ns f# 2.8 Focal Distance 0.7m DI Fuel Injector  Needle Lift: 5V Hold Duration: cf. Table 4-5 Kistler Charge Amplifier  Response: 1pC/psi 20psi/V Time Constant: short, 0.1s Table 4-7 Miscellaneous Equipment Settings  ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 73 4▪8▪7 TEST RUN RANDOMIZATION Each of the eleven test condition that form the CCD test matrix (nine point and three central repeats) was entered in Excel and repeated 51 times. Each row (corresponding to a test condition) was assigned a random number between 0 and 1 and the tests were then completed in ascending random number order. Table 4-8 shows the test order for the capillary injected PSC tests and Table 4-9 those of the insert injected tests.  Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition CAP001 2 0 CAP113 1.75 0 CAP225 1.57 3.53 CAP337 1.75 0 CAP449 1.93 3.53 CAP002 1.57 3.53 CAP114 1.93 3.53 CAP226 1.75 0 CAP338 1.93 -3.53 CAP450 1.93 3.53 CAP003 1.75 0 CAP115 1.57 -3.53 CAP227 1.57 -3.53 CAP339 1.93 -3.53 CAP451 1.5 0 CAP004 1.75 0 CAP116 1.93 3.53 CAP228 1.57 3.53 CAP340 1.57 3.53 CAP452 1.93 3.53 CAP005 1.93 3.53 CAP117 1.57 3.53 CAP229 1.75 0 CAP341 1.75 -5 CAP453 1.75 0 CAP006 1.5 0 CAP118 1.75 0 CAP230 1.5 0 CAP342 1.93 3.53 CAP454 1.75 5 CAP007 1.75 0 CAP119 1.75 5 CAP231 1.75 0 CAP343 1.75 0 CAP455 1.75 5 CAP008 1.5 0 CAP120 1.75 0 CAP232 1.93 3.53 CAP344 1.5 0 CAP456 1.75 0 CAP009 1.75 -5 CAP121 1.93 3.53 CAP233 2 0 CAP345 1.75 -5 CAP457 1.75 5 CAP010 1.75 5 CAP122 1.57 -3.53 CAP234 1.75 5 CAP346 1.75 -5 CAP458 1.75 -5 CAP011 1.75 0 CAP123 2 0 CAP235 1.57 -3.53 CAP347 1.57 3.53 CAP459 1.75 0 CAP012 1.57 3.53 CAP124 1.75 0 CAP236 1.93 -3.53 CAP348 1.57 3.53 CAP460 1.57 -3.53 CAP013 1.75 0 CAP125 1.75 0 CAP237 1.75 -5 CAP349 1.57 3.53 CAP461 1.93 -3.53 CAP014 1.93 -3.53 CAP126 1.5 0 CAP238 1.75 0 CAP350 1.57 3.53 CAP462 2 0 CAP015 1.75 -5 CAP127 1.57 3.53 CAP239 2 0 CAP351 1.75 5 CAP463 1.93 3.53 CAP016 1.93 3.53 CAP128 1.57 -3.53 CAP240 2 0 CAP352 1.75 0 CAP464 1.75 -5 CAP017 1.57 -3.53 CAP129 1.93 -3.53 CAP241 1.5 0 CAP353 1.93 -3.53 CAP465 1.5 0 CAP018 1.57 3.53 CAP130 1.75 0 CAP242 1.75 0 CAP354 2 0 CAP466 1.57 3.53 CAP019 2 0 CAP131 1.5 0 CAP243 1.75 5 CAP355 1.93 -3.53 CAP467 1.57 3.53 CAP020 1.57 -3.53 CAP132 1.5 0 CAP244 1.5 0 CAP356 1.75 0 CAP468 1.5 0 CAP021 1.57 -3.53 CAP133 1.75 0 CAP245 1.75 0 CAP357 2 0 CAP469 1.75 0 CAP022 1.75 0 CAP134 1.93 -3.53 CAP246 1.57 -3.53 CAP358 1.93 3.53 CAP470 1.93 -3.53 CAP023 1.75 -5 CAP135 2 0 CAP247 1.75 5 CAP359 1.93 3.53 CAP471 1.75 -5 CAP024 1.57 -3.53 CAP136 1.57 -3.53 CAP248 1.57 3.53 CAP360 1.57 3.53 CAP472 1.93 3.53 CAP025 1.75 0 CAP137 1.75 5 CAP249 1.75 -5 CAP361 1.75 0 CAP473 1.75 0 CAP026 1.93 -3.53 CAP138 2 0 CAP250 1.5 0 CAP362 1.75 -5 CAP474 1.75 0 CAP027 1.57 -3.53 CAP139 1.75 0 CAP251 2 0 CAP363 1.75 -5 CAP475 1.75 0 CAP028 2 0 CAP140 1.75 -5 CAP252 1.75 0 CAP364 1.75 0 CAP476 1.5 0 CAP029 1.57 -3.53 CAP141 1.75 5 CAP253 1.57 3.53 CAP365 1.75 5 CAP477 1.75 0 CAP030 1.93 3.53 CAP142 1.93 -3.53 CAP254 1.75 5 CAP366 1.93 3.53 CAP478 1.75 0 CAP031 1.75 0 CAP143 1.75 -5 CAP255 1.5 0 CAP367 1.57 3.53 CAP479 1.57 -3.53 CAP032 1.5 0 CAP144 1.75 0 CAP256 1.93 3.53 CAP368 2 0 CAP480 1.75 5 CAP033 1.5 0 CAP145 1.93 3.53 CAP257 1.93 3.53 CAP369 1.93 3.53 CAP481 1.75 0 CAP034 1.75 0 CAP146 2 0 CAP258 2 0 CAP370 2 0 CAP482 1.93 -3.53 CAP035 1.5 0 CAP147 1.75 0 CAP259 1.75 0 CAP371 1.75 0 CAP483 1.75 0 CAP036 1.75 0 CAP148 1.75 0 CAP260 1.93 3.53 CAP372 1.57 3.53 CAP484 2 0 CAP037 1.75 0 CAP149 1.75 0 CAP261 1.93 -3.53 CAP373 1.75 0 CAP485 1.57 3.53 ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 74 Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition CAP038 1.5 0 CAP150 2 0 CAP262 1.75 5 CAP374 1.75 0 CAP486 1.57 3.53 CAP039 1.75 0 CAP151 1.57 -3.53 CAP263 1.75 0 CAP375 2 0 CAP487 1.93 -3.53 CAP040 1.75 0 CAP152 1.75 5 CAP264 1.75 0 CAP376 1.75 5 CAP488 1.75 5 CAP041 1.75 0 CAP153 1.75 5 CAP265 1.75 0 CAP377 1.93 -3.53 CAP489 1.75 0 CAP042 1.93 3.53 CAP154 1.75 -5 CAP266 2 0 CAP378 1.75 -5 CAP490 1.57 -3.53 CAP043 1.57 -3.53 CAP155 1.57 -3.53 CAP267 1.75 0 CAP379 1.75 0 CAP491 1.57 3.53 CAP044 1.75 0 CAP156 1.75 0 CAP268 1.75 0 CAP380 1.93 -3.53 CAP492 1.93 -3.53 CAP045 1.75 0 CAP157 1.57 -3.53 CAP269 1.93 -3.53 CAP381 1.75 0 CAP493 1.75 0 CAP046 1.75 0 CAP158 2 0 CAP270 1.5 0 CAP382 1.57 -3.53 CAP494 1.57 3.53 CAP047 1.75 0 CAP159 1.75 0 CAP271 1.75 -5 CAP383 1.93 3.53 CAP495 2 0 CAP048 1.75 0 CAP160 1.75 5 CAP272 1.93 3.53 CAP384 1.75 5 CAP496 1.93 3.53 CAP049 1.93 -3.53 CAP161 1.57 -3.53 CAP273 1.75 0 CAP385 1.75 0 CAP497 1.5 0 CAP050 1.75 0 CAP162 1.75 5 CAP274 1.75 -5 CAP386 1.57 -3.53 CAP498 1.75 -5 CAP051 1.93 3.53 CAP163 1.57 -3.53 CAP275 1.75 5 CAP387 1.93 -3.53 CAP499 1.75 0 CAP052 1.93 -3.53 CAP164 1.57 -3.53 CAP276 1.75 5 CAP388 1.57 3.53 CAP500 1.75 5 CAP053 1.93 -3.53 CAP165 1.75 -5 CAP277 1.57 -3.53 CAP389 1.5 0 CAP501 1.93 3.53 CAP054 1.57 3.53 CAP166 1.75 0 CAP278 1.75 -5 CAP390 1.75 0 CAP502 1.75 0 CAP055 1.57 3.53 CAP167 1.57 -3.53 CAP279 1.75 5 CAP391 1.93 -3.53 CAP503 1.5 0 CAP056 1.75 0 CAP168 1.5 0 CAP280 1.75 -5 CAP392 1.57 -3.53 CAP504 1.5 0 CAP057 2 0 CAP169 1.75 0 CAP281 1.75 0 CAP393 1.75 5 CAP505 1.57 3.53 CAP058 1.57 3.53 CAP170 1.93 3.53 CAP282 1.75 -5 CAP394 1.93 3.53 CAP506 1.75 -5 CAP059 2 0 CAP171 2 0 CAP283 1.75 0 CAP395 1.5 0 CAP507 1.57 3.53 CAP060 1.57 3.53 CAP172 1.75 0 CAP284 1.5 0 CAP396 1.57 3.53 CAP508 1.57 3.53 CAP061 1.75 -5 CAP173 1.75 0 CAP285 2 0 CAP397 2 0 CAP509 1.57 -3.53 CAP062 1.93 -3.53 CAP174 2 0 CAP286 1.75 -5 CAP398 1.75 0 CAP510 1.75 0 CAP063 1.75 0 CAP175 1.75 0 CAP287 1.75 5 CAP399 1.75 0 CAP511 1.57 3.53 CAP064 1.57 -3.53 CAP176 1.93 -3.53 CAP288 1.75 0 CAP400 1.75 -5 CAP512 2 0 CAP065 1.75 -5 CAP177 2 0 CAP289 1.93 3.53 CAP401 1.75 0 CAP513 1.57 -3.53 CAP066 1.75 5 CAP178 1.93 -3.53 CAP290 1.5 0 CAP402 1.75 5 CAP514 1.75 0 CAP067 1.93 -3.53 CAP179 1.75 5 CAP291 1.57 -3.53 CAP403 1.75 0 CAP515 1.93 3.53 CAP068 1.75 0 CAP180 1.75 0 CAP292 1.93 -3.53 CAP404 1.75 -5 CAP516 1.75 -5 CAP069 1.75 -5 CAP181 1.75 -5 CAP293 1.75 -5 CAP405 1.5 0 CAP517 1.93 -3.53 CAP070 1.75 5 CAP182 1.93 3.53 CAP294 1.57 3.53 CAP406 1.75 0 CAP518 1.5 0 CAP071 1.75 0 CAP183 1.75 -5 CAP295 1.75 0 CAP407 1.75 0 CAP519 1.93 3.53 CAP072 1.93 -3.53 CAP184 1.57 3.53 CAP296 1.57 3.53 CAP408 1.93 3.53 CAP520 1.75 -5 CAP073 1.5 0 CAP185 1.75 0 CAP297 1.57 3.53 CAP409 1.57 -3.53 CAP521 1.75 0 CAP074 1.75 -5 CAP186 1.57 3.53 CAP298 1.75 0 CAP410 1.93 -3.53 CAP522 1.93 -3.53 CAP075 1.57 3.53 CAP187 1.75 -5 CAP299 2 0 CAP411 1.93 -3.53 CAP523 1.75 0 CAP076 1.75 0 CAP188 1.57 3.53 CAP300 1.5 0 CAP412 1.75 0 CAP524 1.75 0 CAP077 1.5 0 CAP189 1.57 -3.53 CAP301 1.75 0 CAP413 1.57 3.53 CAP525 1.75 0 CAP078 1.93 3.53 CAP190 1.75 5 CAP302 1.93 -3.53 CAP414 2 0 CAP526 2 0 CAP079 1.75 5 CAP191 1.75 0 CAP303 1.75 0 CAP415 1.75 0 CAP527 1.57 -3.53 CAP080 2 0 CAP192 1.75 5 CAP304 1.93 -3.53 CAP416 1.93 -3.53 CAP528 1.5 0 CAP081 1.75 -5 CAP193 2 0 CAP305 2 0 CAP417 1.57 3.53 CAP529 1.93 3.53 CAP082 1.75 0 CAP194 2 0 CAP306 1.93 3.53 CAP418 1.75 0 CAP530 2 0 CAP083 1.75 0 CAP195 1.5 0 CAP307 1.75 0 CAP419 1.57 -3.53 CAP531 1.5 0 CAP084 1.75 0 CAP196 1.93 3.53 CAP308 1.75 5 CAP420 1.75 5 CAP532 1.75 0 ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 75 Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition CAP085 1.57 -3.53 CAP197 1.75 0 CAP309 1.57 -3.53 CAP421 1.75 -5 CAP533 1.75 0 CAP086 1.75 5 CAP198 1.93 -3.53 CAP310 1.75 0 CAP422 1.75 0 CAP534 1.75 5 CAP087 1.75 0 CAP199 1.75 0 CAP311 1.75 0 CAP423 1.93 3.53 CAP535 1.75 0 CAP088 1.5 0 CAP200 1.93 3.53 CAP312 1.5 0 CAP424 1.93 -3.53 CAP536 1.93 3.53 CAP089 1.93 3.53 CAP201 2 0 CAP313 1.75 5 CAP425 2 0 CAP537 1.5 0 CAP090 1.75 -5 CAP202 1.93 -3.53 CAP314 1.75 0 CAP426 1.75 0 CAP538 1.93 3.53 CAP091 1.75 -5 CAP203 1.5 0 CAP315 1.75 5 CAP427 1.75 0 CAP539 1.75 -5 CAP092 1.75 -5 CAP204 2 0 CAP316 1.5 0 CAP428 1.57 -3.53 CAP540 1.75 0 CAP093 1.75 0 CAP205 2 0 CAP317 1.75 -5 CAP429 1.57 3.53 CAP541 1.75 5 CAP094 2 0 CAP206 1.75 0 CAP318 1.5 0 CAP430 2 0 CAP542 1.75 0 CAP095 1.75 0 CAP207 1.75 5 CAP319 1.75 0 CAP431 1.93 -3.53 CAP543 1.57 3.53 CAP096 1.75 5 CAP208 1.57 -3.53 CAP320 1.75 5 CAP432 1.93 -3.53 CAP544 1.93 3.53 CAP097 1.93 -3.53 CAP209 1.57 3.53 CAP321 1.75 0 CAP433 1.5 0 CAP545 2 0 CAP098 1.75 -5 CAP210 1.57 -3.53 CAP322 1.75 5 CAP434 1.57 -3.53 CAP546 1.93 -3.53 CAP099 1.75 0 CAP211 1.75 0 CAP323 1.75 0 CAP435 1.57 -3.53 CAP547 1.75 0 CAP100 1.75 0 CAP212 1.57 -3.53 CAP324 1.5 0 CAP436 1.57 3.53 CAP548 1.75 0 CAP101 1.57 -3.53 CAP213 1.93 -3.53 CAP325 1.93 3.53 CAP437 1.75 0 CAP549 1.75 0 CAP102 2 0 CAP214 1.75 5 CAP326 1.75 -5 CAP438 1.93 3.53 CAP550 1.93 3.53 CAP103 1.57 -3.53 CAP215 1.75 0 CAP327 1.57 -3.53 CAP439 1.57 -3.53 CAP551 1.75 0 CAP104 1.5 0 CAP216 1.93 3.53 CAP328 1.5 0 CAP440 1.75 5 CAP552 1.75 0 CAP105 1.5 0 CAP217 1.75 0 CAP329 1.75 0 CAP441 2 0 CAP553 1.93 -3.53 CAP106 1.93 -3.53 CAP218 1.75 -5 CAP330 1.57 3.53 CAP442 1.93 -3.53 CAP554 1.75 0 CAP107 1.75 5 CAP219 1.93 3.53 CAP331 2 0 CAP443 1.5 0 CAP555 1.57 3.53 CAP108 1.75 5 CAP220 1.75 0 CAP332 1.57 -3.53 CAP444 1.75 0 CAP556 1.5 0 CAP109 1.75 0 CAP221 1.93 3.53 CAP333 1.57 3.53 CAP445 1.5 0 CAP557 1.75 -5 CAP110 1.93 -3.53 CAP222 2 0 CAP334 1.75 0 CAP446 1.5 0 CAP558 1.93 3.53 CAP111 1.75 -5 CAP223 2 0 CAP335 1.93 -3.53 CAP447 1.75 5 CAP559 1.93 -3.53 CAP112 1.57 -3.53 CAP224 1.57 3.53 CAP336 1.75 0 CAP448 1.57 3.53 CAP560 1.75 -5             CAP561 1.57 -3.53 Table 4-8 Capillary Injected PSC with DI Experimental Test Order  Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition INS001 1.75 0 INS113 1.57 3.53 INS225 1.75 0 INS337 1.75 0 INS449 1.57 -3.53 INS002 1.57 -3.53 INS114 1.57 -3.53 INS226 1.93 3.53 INS338 1.93 3.53 INS450 1.75 -5 INS003 1.75 -5 INS115 1.75 0 INS227 2 0 INS339 1.5 0 INS451 1.75 5 INS004 1.57 3.53 INS116 1.93 3.53 INS228 1.75 0 INS340 1.75 0 INS452 1.57 3.53 INS005 1.75 0 INS117 1.5 0 INS229 1.93 3.53 INS341 1.75 0 INS453 2 0 INS006 1.57 3.53 INS118 1.75 0 INS230 1.75 0 INS342 1.57 -3.53 INS454 2 0 INS007 1.75 0 INS119 1.93 3.53 INS231 1.75 5 INS343 1.93 -3.53 INS455 1.75 -5 INS008 1.93 3.53 INS120 1.57 -3.53 INS232 1.93 3.53 INS344 1.75 0 INS456 1.75 0 INS009 1.93 -3.53 INS121 1.93 -3.53 INS233 1.5 0 INS345 2 0 INS457 1.75 0 INS010 2 0 INS122 1.57 3.53 INS234 1.75 -5 INS346 1.75 0 INS458 1.75 0 INS011 2 0 INS123 2 0 INS235 1.93 -3.53 INS347 1.75 0 INS459 1.75 5 INS012 1.5 0 INS124 1.75 0 INS236 1.57 -3.53 INS348 1.75 -5 INS460 1.57 -3.53 INS013 1.75 0 INS125 1.75 -5 INS237 2 0 INS349 1.75 5 INS461 1.93 3.53 INS014 1.75 -5 INS126 1.75 5 INS238 1.57 3.53 INS350 1.75 -5 INS462 1.57 -3.53 INS015 1.57 -3.53 INS127 1.75 0 INS239 1.57 -3.53 INS351 1.75 0 INS463 1.75 0 ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 76 Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition INS016 1.75 0 INS128 1.93 3.53 INS240 1.75 0 INS352 1.5 0 INS464 1.93 -3.53 INS017 1.93 3.53 INS129 1.75 0 INS241 1.75 0 INS353 1.75 0 INS465 1.75 0 INS018 1.75 0 INS130 2 0 INS242 1.75 0 INS354 1.75 0 INS466 1.75 0 INS019 1.75 0 INS131 1.5 0 INS243 1.75 0 INS355 1.57 3.53 INS467 1.57 3.53 INS020 1.93 3.53 INS132 1.75 5 INS244 1.75 5 INS356 1.93 3.53 INS468 2 0 INS021 1.57 3.53 INS133 1.75 0 INS245 1.57 -3.53 INS357 1.75 0 INS469 1.75 -5 INS022 1.93 -3.53 INS134 2 0 INS246 1.93 3.53 INS358 1.57 3.53 INS470 1.75 -5 INS023 1.57 -3.53 INS135 1.93 -3.53 INS247 1.75 0 INS359 1.75 0 INS471 1.93 3.53 INS024 1.75 0 INS136 1.75 0 INS248 1.75 0 INS360 1.57 3.53 INS472 2 0 INS025 1.5 0 INS137 1.75 0 INS249 1.57 3.53 INS361 1.75 0 INS473 1.75 0 INS026 1.75 0 INS138 2 0 INS250 1.75 5 INS362 1.93 3.53 INS474 1.57 3.53 INS027 1.75 0 INS139 1.57 3.53 INS251 2 0 INS363 1.75 5 INS475 1.75 0 INS028 1.93 -3.53 INS140 1.5 0 INS252 1.93 -3.53 INS364 1.57 3.53 INS476 1.93 -3.53 INS029 1.57 3.53 INS141 1.75 0 INS253 1.5 0 INS365 1.75 -5 INS477 1.75 0 INS030 1.75 5 INS142 1.75 0 INS254 1.75 -5 INS366 2 0 INS478 1.75 5 INS031 1.75 0 INS143 1.93 3.53 INS255 1.93 -3.53 INS367 1.57 -3.53 INS479 1.75 0 INS032 1.93 -3.53 INS144 1.57 3.53 INS256 1.5 0 INS368 1.93 3.53 INS480 1.57 -3.53 INS033 1.57 -3.53 INS145 1.57 -3.53 INS257 2 0 INS369 1.75 0 INS481 1.57 3.53 INS034 1.75 5 INS146 2 0 INS258 1.93 -3.53 INS370 1.57 3.53 INS482 1.57 3.53 INS035 1.75 0 INS147 1.75 -5 INS259 1.75 0 INS371 1.75 -5 INS483 1.75 0 INS036 1.57 -3.53 INS148 1.75 0 INS260 2 0 INS372 1.75 -5 INS484 1.93 -3.53 INS037 1.75 0 INS149 2 0 INS261 1.75 -5 INS373 2 0 INS485 1.75 5 INS038 1.5 0 INS150 1.5 0 INS262 1.75 5 INS374 1.5 0 INS486 1.93 -3.53 INS039 1.75 0 INS151 1.5 0 INS263 1.93 3.53 INS375 1.75 5 INS487 1.75 0 INS040 1.75 0 INS152 1.75 0 INS264 1.75 0 INS376 1.93 3.53 INS488 1.75 -5 INS041 1.75 5 INS153 1.93 3.53 INS265 1.75 0 INS377 1.57 3.53 INS489 1.75 0 INS042 1.75 0 INS154 1.93 3.53 INS266 1.57 -3.53 INS378 1.5 0 INS490 2 0 INS043 1.5 0 INS155 1.75 0 INS267 1.75 5 INS379 1.75 -5 INS491 1.75 0 INS044 1.75 5 INS156 1.75 5 INS268 1.75 0 INS380 1.57 3.53 INS492 1.93 -3.53 INS045 1.75 0 INS157 1.75 5 INS269 1.75 0 INS381 1.57 -3.53 INS493 1.57 -3.53 INS046 1.57 3.53 INS158 1.57 -3.53 INS270 1.75 -5 INS382 1.75 5 INS494 1.93 3.53 INS047 1.57 -3.53 INS159 1.75 5 INS271 1.93 -3.53 INS383 1.5 0 INS495 1.57 3.53 INS048 1.75 -5 INS160 1.57 -3.53 INS272 1.57 -3.53 INS384 1.75 0 INS496 1.75 0 INS049 1.75 0 INS161 1.93 -3.53 INS273 1.75 0 INS385 1.75 5 INS497 1.93 3.53 INS050 1.57 -3.53 INS162 1.93 3.53 INS274 1.93 -3.53 INS386 1.57 3.53 INS498 1.75 0 INS051 1.75 0 INS163 1.93 -3.53 INS275 1.57 3.53 INS387 1.57 -3.53 INS499 1.75 5 INS052 1.93 3.53 INS164 1.93 3.53 INS276 1.93 -3.53 INS388 1.75 0 INS500 1.75 5 INS053 1.75 5 INS165 1.75 0 INS277 1.75 0 INS389 1.5 0 INS501 1.75 -5 INS054 1.75 5 INS166 1.75 0 INS278 1.75 5 INS390 1.75 0 INS502 1.57 -3.53 INS055 1.75 5 INS167 1.75 0 INS279 1.75 5 INS391 1.75 0 INS503 1.75 0 INS056 1.75 0 INS168 1.75 0 INS280 1.93 3.53 INS392 1.5 0 INS504 1.57 -3.53 INS057 1.57 -3.53 INS169 1.93 3.53 INS281 1.5 0 INS393 1.93 -3.53 INS505 1.93 -3.53 INS058 1.93 -3.53 INS170 2 0 INS282 1.75 0 INS394 1.75 0 INS506 1.57 3.53 INS059 1.5 0 INS171 2 0 INS283 1.57 -3.53 INS395 1.75 -5 INS507 1.93 3.53 INS060 1.93 -3.53 INS172 1.75 -5 INS284 1.75 -5 INS396 1.5 0 INS508 1.75 -5 INS061 1.57 -3.53 INS173 1.57 3.53 INS285 1.75 0 INS397 1.75 0 INS509 1.57 3.53 INS062 2 0 INS174 1.75 0 INS286 1.75 0 INS398 2 0 INS510 1.75 0 ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 77 Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition INS063 1.75 0 INS175 1.75 5 INS287 1.75 5 INS399 1.75 -5 INS511 2 0 INS064 1.75 0 INS176 1.75 5 INS288 1.5 0 INS400 1.75 0 INS512 1.75 -5 INS065 1.57 -3.53 INS177 1.75 0 INS289 1.57 3.53 INS401 1.75 5 INS513 1.75 -5 INS066 1.93 -3.53 INS178 1.75 0 INS290 2 0 INS402 1.93 3.53 INS514 1.75 0 INS067 1.5 0 INS179 2 0 INS291 2 0 INS403 1.5 0 INS515 1.75 0 INS068 1.57 3.53 INS180 2 0 INS292 1.57 3.53 INS404 1.5 0 INS516 1.75 -5 INS069 1.75 -5 INS181 1.5 0 INS293 1.93 3.53 INS405 1.75 -5 INS517 1.75 -5 INS070 1.75 0 INS182 2 0 INS294 2 0 INS406 1.5 0 INS518 1.93 3.53 INS071 2 0 INS183 2 0 INS295 2 0 INS407 1.57 -3.53 INS519 1.75 -5 INS072 1.75 0 INS184 1.75 5 INS296 1.75 5 INS408 1.75 -5 INS520 1.93 3.53 INS073 1.75 5 INS185 1.93 -3.53 INS297 1.75 5 INS409 1.5 0 INS521 1.93 -3.53 INS074 1.57 -3.53 INS186 1.75 0 INS298 1.75 5 INS410 1.93 -3.53 INS522 1.93 3.53 INS075 1.93 3.53 INS187 1.75 -5 INS299 2 0 INS411 1.93 -3.53 INS523 2 0 INS076 1.75 5 INS188 1.75 -5 INS300 1.93 -3.53 INS412 1.93 3.53 INS524 1.75 0 INS077 1.93 -3.53 INS189 1.57 3.53 INS301 1.5 0 INS413 1.75 0 INS525 1.75 -5 INS078 1.75 0 INS190 1.5 0 INS302 1.75 0 INS414 1.5 0 INS526 1.75 0 INS079 1.57 3.53 INS191 1.93 3.53 INS303 1.75 0 INS415 1.57 -3.53 INS527 1.93 3.53 INS080 1.75 0 INS192 1.75 -5 INS304 1.75 5 INS416 1.93 -3.53 INS528 1.93 -3.53 INS081 1.93 -3.53 INS193 1.75 0 INS305 1.57 -3.53 INS417 1.75 0 INS529 1.75 -5 INS082 1.57 3.53 INS194 1.75 5 INS306 1.57 3.53 INS418 1.75 0 INS530 1.75 0 INS083 1.57 3.53 INS195 1.75 5 INS307 1.75 5 INS419 1.93 3.53 INS531 1.93 -3.53 INS084 1.93 -3.53 INS196 1.75 0 INS308 1.5 0 INS420 1.75 -5 INS532 1.93 -3.53 INS085 1.75 5 INS197 1.75 0 INS309 1.93 -3.53 INS421 1.75 0 INS533 1.75 0 INS086 1.57 3.53 INS198 2 0 INS310 1.93 -3.53 INS422 1.93 -3.53 INS534 1.57 3.53 INS087 1.5 0 INS199 1.5 0 INS311 1.93 3.53 INS423 1.75 0 INS535 1.57 3.53 INS088 1.75 0 INS200 1.75 -5 INS312 1.75 0 INS424 1.93 3.53 INS536 1.75 0 INS089 2 0 INS201 1.57 3.53 INS313 1.93 -3.53 INS425 2 0 INS537 1.93 3.53 INS090 1.57 3.53 INS202 1.75 0 INS314 1.75 0 INS426 1.93 -3.53 INS538 1.75 0 INS091 1.93 3.53 INS203 1.93 -3.53 INS315 1.57 -3.53 INS427 1.57 -3.53 INS539 1.75 0 INS092 1.93 -3.53 INS204 2 0 INS316 1.75 5 INS428 1.75 0 INS540 1.93 3.53 INS093 2 0 INS205 1.75 -5 INS317 1.75 -5 INS429 1.57 -3.53 INS541 2 0 INS094 1.75 0 INS206 1.57 -3.53 INS318 1.75 0 INS430 1.75 -5 INS542 1.57 3.53 INS095 1.75 -5 INS207 1.75 0 INS319 1.75 0 INS431 1.75 0 INS543 1.75 0 INS096 1.75 5 INS208 1.93 -3.53 INS320 1.5 0 INS432 1.75 5 INS544 1.57 -3.53 INS097 1.93 -3.53 INS209 1.93 3.53 INS321 1.5 0 INS433 1.75 0 INS545 1.57 -3.53 INS098 1.57 -3.53 INS210 1.57 3.53 INS322 1.75 0 INS434 1.75 5 INS546 1.5 0 INS099 1.57 -3.53 INS211 1.75 0 INS323 2 0 INS435 1.57 3.53 INS547 1.5 0 INS100 2 0 INS212 1.75 0 INS324 2 0 INS436 1.57 -3.53 INS548 2 0 INS101 1.75 0 INS213 1.75 -5 INS325 1.75 -5 INS437 1.57 -3.53 INS549 1.5 0 INS102 1.75 0 INS214 1.75 0 INS326 1.5 0 INS438 1.93 -3.53 INS550 1.57 -3.53 INS103 1.75 0 INS215 1.57 3.53 INS327 1.57 3.53 INS439 1.75 5 INS551 1.75 0 INS104 1.93 3.53 INS216 1.57 3.53 INS328 1.5 0 INS440 1.5 0 INS552 1.5 0 INS105 1.75 0 INS217 1.5 0 INS329 1.93 -3.53 INS441 1.75 0 INS553 1.93 3.53 INS106 1.5 0 INS218 2 0 INS330 1.75 -5 INS442 1.75 -5 INS554 1.5 0 INS107 1.75 0 INS219 2 0 INS331 1.75 0 INS443 1.75 0 INS555 1.93 3.53 INS108 1.57 -3.53 INS220 1.75 5 INS332 1.57 3.53 INS444 1.5 0 INS556 1.93 -3.53 INS109 1.57 -3.53 INS221 1.75 0 INS333 1.5 0 INS445 1.57 3.53 INS557 1.5 0 ▪ Chapter 4 ▪ Experimental Setup & Procedures ▪ 78 Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition Run ID Bulk Charge Timing Condition INS110 1.93 3.53 INS222 1.75 -5 INS334 2 0 INS446 1.57 -3.53 INS558 1.75 0 INS111 1.93 3.53 INS223 1.75 -5 INS335 2 0 INS447 1.57 -3.53 INS559 1.93 -3.53 INS112 1.5 0 INS224 1.75 0 INS336 1.93 3.53 INS448 1.75 0 INS560 1.75 0             INS561 1.57 3.53 Table 4-9 Insert Injected PSC with DI Experimental Test Order     79  Chapter 5   REPEATABILITY & CALIBRATION TESTS   5▪1 INTRODUCTION To ensure the quality of the data collected form the LIF system there are a number of supporting assessments that must be made of the experimental system. Chapter 5 details the procedures followed and the findings of a series of experiments that were undertaken to assess a range of factors: the repeatability of the camera timing, the spatial and temporal variation in the camera CCD response and the laser beam profile. Theoretical presentations are made that demonstrate the fluorescent response to be linear with laser power, while experimental findings are used to show that any fluorescence from natural gas may be discounted when compared to the signal produced from acetone. An assessment of the polytropic exponent for the RCM compression stroke is made and high speed camera footage of the combustion in the RCM cylinder demonstrates the lack of in cylinder turbulence. Finally, the experimental calibration curves for the fuel injectors, acetone seeder and acetone fluorescent response are presented.  ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  80 5▪2 CAMERA TRIGGER REPEATABILITY The triggering of the iCCD camera was checked for its repeatability with respect to piston position. The RCM supply pressure was regulated to 80psi while the camera timing was randomly selected to be 4300, 4800 or 5800 shaft encoder counts. Each image was binarised and the piston edge identified by thresholding; piston position could then be inferred from to total black pixel count. For each set of 101 images a running mean position was calculated and the standard deviation evaluated accordingly. Finally, through imaging a calibration grid, the pixel counts were converted into a ‘true’ position in mm. The results for three different timings show that the camera trigger may be deemed repeatable to within 0.9CAD.   Figure 5-1 Mean Piston Position and Standard Deviation of Position for 4300, 4800 and 5800 Camera Trigger Count Timing ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  81 5▪3 ICCD CALIBRATION The iCCD camera is known to add noise and spatial discrepancy (spatial noise) to images; each was evaluated according to the procedures detailed below. All images were taken with a Cooke Corporation DiCam Pro iCCD camera with a Nikor AF Micro lens, 105mm, 1:2.8D, set to f#2.8 and camera intensifier gain of 99.5% (lowest noise setting for the MCP). 5▪3▪1 SPATIAL VARIATION OF ICCD INTENSITY COUNT A 15 by 12 array of diffuse blue LED’s (λpeak = 428nm and λdominant = 466nm, 50% intensity at ±30 0) uniformly spaced at 0.3” was used to form a flat field light source. The light emitted was imaged at right angles to a set of theatrical screening which had the LED array behind it. An exposure of 350µs was chosen to match the peak of the image intensity histogram to the middle of the camera’s dynamic range. Each exposure was repeated 256 times and then arithmetically averaged (using the camera’s CamWare imaging software) to provide the final image. Figure 5-2 shows two representations of the same control image. The left shows the basic image across the full intensity range, while that on the right shows the same image with a range crop applied to the intensity levels.        (a)       (b)  Figure 5-2  iCCD Flat Field Image with Mean Intensity at 26 Bits, With (right) and Without (left) Range Cropping  A clear variation in the intensity of the image is visible in the right hand image of Figure 5-2(b). The peak intensity is about 11.8±2.4% (95% confidence) above the field mean, with the minimum intensity correspondingly low by about 11.6±1.9% (95% confidence). Following discussions with the camera manufacturer the variation is believed to come from manufacturing variance in the MCP of the camera intensifier. There is no on-board correction procedure and the variation must be corrected for in post- processing.  ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  82 5▪3▪2 NOISE FLOOR FOR ICCD The noise found in any image produced from a CCD, intensified or not, can be attributed to a number of sources. Spatial inconsistency (i.e. noise) is directly related to the wavelength of the incident radiation [369], but in the visible spectrum the wavelength of the system noise is substantially shorter than that of the incident light, allowing the spatial noise attributable to incident radiation frequency to be overlooked. However, there are three primary noise sources on the camera/CCD which cannot be overlooked. Readout noise occurs directly from transferring charge from one well to the next when reading the charge off the chip to the serial register(s), and scales linearly with readout rate. Secondly, the Johnson-Nyquist noise (also called dark or thermal noise) is attributable to the thermal excitation of the electrons on the CCD, and is directly proportional to both the chip temperature and readout time. The intensifier further adds to the system noise because of the amplification it is designed to provide and its non-ideal quantum efficiency. The individual effects of each of these factors cannot be separated and together form the noise floor for the camera.   A series of 51 dark images were taken, each of which was the arithmetic average of 256 exposures (made using the camera’s CamWare image control software). These images were made in a completely darkened laboratory at an exposure of 350µs. The imaging surface was that used in the spatial variation study (cf. §5▪3▪1).  The dark field image for the camera used in this study was found to have a full range of 8 pixel counts, with a mean of 87. Figure 5-3 shows the mean pixel count against the standard deviation for the series of 51 dark images. The broken lines indicate the mean image count and mean standard deviation for the complete image set. The figure indicates that the higher the average dark field count the higher the noise level (standard deviation) in the image. The random nature of the noise floor cannot be corrected for, and will be taken to have a value of 87±0.06 (95% confidence) counts for the purpose of all image processing.   ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  83  Figure 5-3  Mean and Standard Deviation for the iCCD Dark Field 5▪3▪3 TEMPORAL VARIATION In order to assess properly the noise floor of the iCCD, any temporal variation in the imaging procedure must be quantified. Using the approach outline by Rothamer & Gandhi [369] the temporal variation of the camera flat field was studied. A series of 51 flat field images was taken over an eight hour period. As discussed in §5▪3▪1 the flat field image taken by the camera is not flat. To account for this a subregion was identified around the image maximum. The subregion was defined to be 200 × 200 pixels, giving a total area of 40000 pixels2. Figure 5-4 shows plots for the central row of pixels in the x (left) and y (right) directions for the subregion.  Figure 5-4  Central Row (left) and Column (right) for iCCD Flat Field Subregion  Both plots show good uniformity, however, in processing the images for any time variation, the uniformity of the image is no longer assured. The subregion in each image was normalised against the mean for the entire image set. This provided a time normalised result for each image. Each subregion, was ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  84 subsequently divided into four further regions, for which each sub-subregion mean was normalised against the mean for the other three. This provided a spatially normalised result for each image. The time and spatially normalised mean values were plotted against each other to form Figure 5-5.   The four rows of data points make it apparent that the flat field is not perfectly spatially uniform. Error bars are not shown since the CI95% values were 5.15×10 -5 and 3.85×10-5 for the abscissa and ordinate values respectively. Without making the subregion insignificantly small this variation cannot be suppressed. This does not, however, preclude the use of Figure 5-5 in evaluating the temporal variation in the flat field.  Figure 5-5  Mean Temporal-Spatial Correlation Plots for iCCD Flat Field  The confidence interval (for any individual sub-subregion) is of the order CI95% ±0.15%, which corresponds to an SNR of ~1350. This value is far beyond that achievable in the shot-noise-limited case (which corresponds to about ~100 [369]). The order of magnitude difference between the temporal and shot-noise SNR allows any temporal variation in camera behaviour to be disregarded. 5▪4 FLUORESCENT SIGNAL CHECKS 5▪4▪1 SIGNAL LINEARITY & DISSOCIATION CHECKS When trying to ascribe natural gas concentrations to the intensity count recorded by the iCCD, it is assumed that the signal from the acetone is proportional to both laser power and acetone concentration. For this to hold, population inversion of the excited acetone must be avoided to preclude saturation of the signal. This is achieved when [196] Ni<<(στ) -1 Eq. 5.1 ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  85 Assuming a fluorescent lifetime of τflu~1.1x10 -9s [306] and an absorption cross-section of σabs~2.3x10 -20 cm2 at 248nm [196, 305]; Ni≈3.95x10 28 photons/cm2s. Each photon has an energy given by  Eq. 5.2 which, for a 20ns pulse, results in a required intensity of  ~633 J/cm2 for signal saturation. However, the laser used in this study provides 166 J/cm2 at maximum power, avoiding signal saturation. Further, dissociation of the acetone must be avoided. The lower dissociation limit for acetone is 86 kcal/mol [336], or ~360kJ/mol. Yet the current laser provides 400mJ maximum energy, which equates to 42kJ/mol at 1% acetone v/v at TDC- well below the dissociation limit. 5▪4▪2 THE FLUORESCENCE OF NATURAL GAS The odour markers included in commercial natural gas are most commonly based on a series of compounds called mercaptans (sulphurous equivalents to alcohols). THT (tetrahydrothiophene, (CH2)4S) is also commonly used and is known to fluoresce significantly in the UV, and has consequently been used as a fuel tracer [370], though this is eschewed in the current study. The composition of natural gas varies geographically and with the time of year. Vancouver natural gas has an approximate composition as shown in Table 5-1. Depending upon the exact sulphur content of the gas blend, 8-24 mg/m3 of marker are added of which 65% is t-butyl mercaptan, (CH3)3CSH, and the reminder methyl-ethyl sulphide, (CH3)4S; both of which are UV transparent13.   Compound Chemical Formula Mole Fraction % Molecular Mass (g/mol) Gas Constant (J/kg.K) Ratio of Specific Heats Methane CH4 94.6643% 16.04 518.33 1.304 Ethane C2H6 1.8016% 30.07 276.49 1.187 Propane C3H8 0.4236% 44.10 188.53 1.27 Butane C4H10 0.1589% 58.12 143.05 1.094 Pentane C5H12 0.0461% 72.15 115.23 1.07 Hexane C6H14 0.0076% 86.17 96.48 1.06 Heptane C7H16 0.0000% --- --- --- Octane C8H18 0.0000% --- --- --- Nitrogen N2 0.5975% 28.01 296.82 1.40 Carbon Dioxide CO2 0.1698% 44.01 188.91 1.289 Table 5-1 MicroGC Natural Gas Composition Results14  To ensure this UV transparency a series of tests where conducted to compare the fluorescent signal obtained from a natural gas and acetone-doped natural gas. Using the RCM, 31 images were taken of homogeneous natural gas charges and 31 natural gas charges seeded with 1% v/v acetone. Arithmetic mean images were created and normalized against a similar mean image for air. Figure 5-6(a) shows the                                                        13 Thanks must go to Jim Cox of Terasen Inc. for the information on odour markers in BC gas. 14 Results from Westport Innovations average of January to July 2009  ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  86 natural gas image normalized against an air charge, while Figure 5-6(b) shows the air normalised gas/acetone mixture.  (a)      (b) Figure 5-6  Air Normalised Images of Fluorescent Intensity from Natural Gas (left) and Natural Gas with 1% Acetone v/v (right).  The natural gas images produce a signal ~20% stronger than air, which may be attributed to poorly excited LIF of various compounds, or inelastically scattered light. The gas/acetone mixture has a signal ~90% stronger than air, significantly higher than that of the unseeded natural gas. The strong signal from the inelastic scattering of the laser light, which causes significant spatial noise, is evident. From the data presented it is not possible to fully discount the signal produced by the natural gas, but given that it is ~4.5 times lower than that produced by the acetone doped charge, and of an intensity which is within the system noise it is possible to attribute the LIF signal in all future images to the acetone/methane mixture present which is dominated by that of the acetone. 5▪4▪3 DISCOUNTING RAYLEIGH SCATTERING In any laser diagnostics system the Rayleigh signal dwarfs those of Raman and LIF. However, the Stokes shift of the latter two allows easy filtering of the Rayleigh signal. Acetone fluoresces in the blue at ~350- 550nm, which is 100-300nm away from the excitation wavelength. Correspondingly, rejection of elastically scattered laser light of the Rayleigh signal can be achieved with a commercial high grade optical UV filter. To ensure that this is the case a series of 31 images were taken of a homogenous air charge and 31 more of the same charge with a natural gas injection event. There should be no discernable difference in the recorded images. Following a similar procedure to that outlined for the natural gas fluorescence checks the two image sets were compared, Figure 5-7. The injection event occurred in the ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  87 bottom left of the image and extended to the top right; given that no discernable structure is apparent Rayleigh scattering may be discounted.   Figure 5-7  Unseeded Direct Injection Event Normalised Against Quiescent Air Charge 5▪5 PRESSURE TRANSDUCER CALIBRATION A piezoelectric dynamic pressure transducer is used to capture the pressure in the RCM cylinder. The PCB Electronics 112A dynamic pressure transducer was installed in a small volume which has a hand operated ball valve on the inlet and outlet. The transducer output was amplified with a Kistler 5051B charge amplifier set at 1pC/psi and a scale of 300 psi/V. A regulated 2000psi nitrogen supply was connected to the inlet, which provided the required pressure step when the inlet valve was opened. The output from the charge amplifier was recorded on a Tektronix 2024B digital oscilloscope. The oscilloscope was also used to analyse the signal for peak-to-peak voltage change. Fifty-one pressurisations and depressurisations were performed and the transducer/amplifier response recorded. The results showed a mean voltage change of 7.4V which corresponds to 3.7mV/psi. It must be noted however that the following effects cannot be accounted for (in order of significance):  The precision of the supply pressure;  The effects of the dead volume, pressure wave formations and shape on valve opening;  Thermal considerations such as charge cooling upon valve actuation;  Case over/under torquing;  Transducer linearity. ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  88 5▪6 POLYTROPIC EXPONENT CALCULATION When considering the compression stroke of an ICE the process is assumed to be isentropic, an assumption which is well established in its validity. However, the process is far from reversible or adiabatic. These factors can be taken into account by evaluating the polytropic exponent of the working fluid, the value of which will be different to the theoretical one. The evaluated polytropic exponent can then be used for subsequent calculations.  The rapid compression machine was filled with an atmospheric air charge at ambient pressure. The machine was then fired and the cylinder pressure trace recorder with the calibrated pressure transducer. The charge signal from the transducer was amplified and converted to a voltage signal which along with, crank position and index pulse, was recorded on a Tektronix 1024B oscilloscope. This voltage data was subsequently converted to a pressure reading (in psi), while the index pulse and crank position were used to calculate the cylinder volume (in cu. in.) for each data point. Following the method outlined by McCullough [371] the slope of a log-log plot of pressure vs. volume provided the polytropic exponent with heat and pressure losses accounted for. The data presented below are an average of 31 separate runs. The procedure outlined was also repeated identically for a stoichiometric CH4/atmospheric air mixture.     (a)           (b) Figure 5-8  Polytropic Exponent Plots for Air (left) and Stoichiometric CH4 Mixture (right)  Figure 5-8(a) clearly shows a strong linear relationship between the natural-logarithms of pressure and volume. As would be expected [371] the peak pressure is seen to be advanced of TDC due to heat loss and leakage. A least squares fit (R2 = 0.9936) of the data suggest a polytropic exponent of 1.26 (10% lower than the true value) which is reasonable when compared to the only other known published value of 1.24 ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  89 [372]. Based on a least squares fit (R2 = 0.9064) to the stoichiometric data, Figure 5-8(b), the polytropic exponent was calculated to be 1.14, or 13.6% below the theoretical value for the mixture. Although not as good as that of acetone the effect upon the experimental results is expected to be small since the exponent value is use to establish the state within the cylinder, and is not used directly in the fuel concentration specification. 5▪7 LASER SHEET PROFILE CORRECTION The beam profile of an excimer laser is known to be approximately Gaussian longitudinally and top-hat transversely [152]. The transverse uniformity of the beam and small width of the light sheet ensure a well defined interrogation volume in one dimensions, however the Gaussian profile of the beam must be corrected for in the second dimension. The RCM cylinder was filled with a pure air charge and, without any cylinder motion, injected with 250 3ms injections from the DI injector (at 62mg/injection). The resultant homogeneous charge of natural gas with 1% v/v acetone vapour was imaged with 31 consecutive shots of the laser. Each image was corrected for background light intensity and had a central region across the image x-direction 10 pixels wide investigated. Each subregion had pixel values averaged for each column and this was repeated for each image. The values for each image were then averaged themselves to form a representative profile of the laser sheet intensity. The result of this process is shown in Figure 5-9.  Figure 5-9  Laser Sheet Profile; Recorded (•) and with a 4th Order Fit (-)  The recorded data had a fourth order fit applied (since it offered better conditioning than a double Gaussian fit) with the coefficients listed in Table 5-2. ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  90 Fit x4 x3 x2 x1 x0 Based on Pixel Value 8.4×10-7 -6.9×10-4 0.1326 5.5419 1342.9 Based on Sheet Position (mm) 2.2×10-4 -0.044 2.0883 23.220 1348.5 Table 5-2 Least Squares Fit Coefficients for Laser Sheet Profile 5▪8 OPTICAL DISTORTION CORRECTION The optical distortion caused by the emitted light passing through the curved surface of the RCM bore mustbe be corrected for. A calibration grid comprising alternate black and white squares measuring 2×2mm was placed in the RCM cylinder and imaged. On the cylinder centreline, where no distortion is expected, the number of pixels per mm can be calibrated. The number of pixels that comprise each alternate black and white square on a calibration grid can be counted and compared to the value on the centreline. The distortion factor, δ, can then be defined by, Eq. 5.3   Eq. 5.3 Figure 5-10 shows the distortion factor for each set of 2×2mm grids starting at the cylinder centreline and progressing towards to the cylinder bore. The ‘steps’ occur because the 2×2mm grids can only correspond to integer pixel dimensions. The curve in Figure 5-10 corresponds to a sine curve over the interval 2.25 – π radians. The first value in this range matches the distortion factor to the experimental value at the cylinder bore, while the second value specifies zero distortion at the centreline. The agreement between the calculated distortion factor and the experimental value is seen to be good.   Figure 5-10 Cylinder Optical Distortion  Images are then corrected by binning the image into 2×2mm blocks and stretching them in a piece-wise fashion to the corrected size (according to Eq. 5.4), with inserted pixel values interpolated from those on either side of it. ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  91  Eq. 5.4 The correction is made in the radial direction only and pixel dimensions rounded to the nearest integer value. This approach leads to images that are 832 pixels tall (when no on-chip camera binning is used). The calibrated pixel size (at the cylinder centreline) is 0.125mm, yielding a corrected image size of 104mm, which is the cylinder bore. 5▪9 PIXEL INTENSITY AIR-FUEL-RATIO CALIBRATION The RCM seeder and premixer were used to generate acetone doped mixtures of natural-gas and air. Mixtures were prepared according to fuel partial-pressures, which were calculated according to the composition given in Table 5-1. Homogeneous charges were prepared at relative air-fuel ratios of: 0 (pure fuel), 0.3, 0.5, 1, 1.5, 2 and ∞ (pure air). The RCM was then filled with the appropriate mixture and a homogeneous charge image collected. Mixtures were prepared to allow batches of fives runs; a total of 51 images were taken for each mixture concentration. All collected images were corrected for background noise, CCD chip-spatial variation, laser power variation, and sheet intensity variation. The corrected images were then averaged and a sub-region, away from reflections, spatially averaged to obtain an intensity count for the given mixture RAFR. From the signal linearity checks of §5▪4▪1 it is axiomatic that that signal intensity should be linear with respect to acetone number count. Assuming fully homogenised seeding the fuel number count can be used as a proxy for acetone number count. The fuel normalised acetone number count  was plotted against the fluorescent intensity. The result is Figure 5-11, which is reassuringly linear (R2 = 0.995).   ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  92 Figure 5-11  Normalised Fuel Number Count vs Signal Intensity  This normalised calibration curve based on number count is equated with relative air fuel ratio as follows. By taking the quotient of the perfect gas equation of state for the air and fuel in the RCM cylinder, the total pressure of air within the mixture can be expressed Eq5.5  Eq. 5.5 Noting that the mass quotient of Eq5.5 is the AFR and given that the relative air-fuel ratio (RAFR) is (Eq5.6):  Eq. 5.6 it is possible through substitution and simple manipulation to yield the calibration equation of Eq5.7,which can be evaluated fully since the ratio Ptotal/Pfuel (equal to Nfuel/Ntotal) is given by the expression for the calibration line of Figure 5-11 and all other quantities are known (cf. Table 5-3)   Eq. 5.7  Property Value AFRstoic 16.911 Molar AFR 9.5098 Molecular Weight 16.2878 g/mol Specific Gas Constant 501.7 J/kgK Ratio of Specific Heats 1.2644 H:C ratio 3.927 CoV Methane 0.761% CoV AFR 0.234% Table 5-3 Mixture Properties Based on Data in Table 5-1 and Original GC Data 5▪10 IMAGE BINNING A comparison of image SNR was made between an un-binned, a 2×2, and 4×4 binned image to assess whether the loss in spatial resolution associated with binning is outweighed by the increased signal-to- noise ratio. Homogeneous doped natural gas/air charges were prepared as they were for the air-fuel ratio calibration tests of §5▪9. On-chip binning of the CCD wells allows for increased pixel intensity (larger optical collection area) without increased camera-noise, while the probabilistic nature of photon detection by CCD’s is well described by Poisson statistics. For an intensified image the signal-to-noise ratio is given ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  93 by Eq5.8 [373] where I and I0 are the signal intensity and camera noise floor respectively and the other parameters are as listed in Table 5-4.  Eq. 5.8  Parameter Value Photocathode Efficiency, η 0.4 Intensifier Noise Factor, κ 1.03 Intensifier Gain 99.5 Table 5-4 DiCam Pro iCCD Noise Parameters  The unbinned pixel corresponded to physical dimensions 0.125 × 0.125 mm, and provided prohibitively poor signal levels above the camera-noise floor, such that only the a 2×2, and 4×4 binned images were assess for their SNR. The SNRs for the eight tested concentrations are as shown in Figure 5-12 (low signal strength detail in inset).  Figure 5-12 Signal-to-Noise Ratio for 2×2 and 4×4 On-Chip Binning at Different Signal Intensities & Maximum SNR Curve  In Poisson statistics the mean and variance are equal and as such, in the shot-noise limit the SNR tends to (curve shown in Figure 5-12), but it is apparent from Eq5.8 that the shot-noise limited case for an intensified system  will have an SNR reduced from its maximum by , approximately the difference observed at the highest signal intensity. The range of intensities from the fuel concentrations shows that all LIF images will be in transition from system to shot-noise limited, depending upon the fuel ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  94 concentration. This informed the decision to take all images with a 4×4 on-chip bin to maximise the image SNR without compromising spatial resolution or causing image saturation (the result of 8×8 binning). 5▪11 SEEDER CHARACTERISATION The seeder used to dope acetone into the methane fuel supply is described in §4▪5 and in detail in Appendix A. The two supplies to the seeder were set so that the flow through the acetone accounted for 0, 25, 50, 75 or 100% of the total seeder gas flow. Three repeats were carried out for each flow condition, with the setting chosen at random. The seeder was allowed to settle to a steady state, with 15 minutes (1.5 residence times) elapsing between samples. The samples were collected in 1 litre Tedlar® bags with two venting samples taken prior to each data sample. The 15 samples were sent for gas chromatograph and mass spectrometry analysis at PowerTech Laboratories Inc., Surrey, BC, Canada. Calibration samples were prepared by PowerTech Labs Inc.  The GC/MS results of PowerTech Labs Inc. report, project#: 15808-43, are presented Figure 5-13. A linear fit provides a regression co-efficient, R2, of 0.9553. A lack of knowledge about the experimental procedures and errors precludes the specification of error bars 15.                                                         15 Thanks must go to Ed Hall of PowerTech Labs for his help in completing the analysis. ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  95 Figure 5-13  Acetone Seed Concentrations 5▪12 FLOW ENVIRONMENT IN RCM CYLINDER The premise of the RCM is that it provides engine-like cylinder conditions with a simplified flow field. The lack of valve system ensures that any turbulence generation is from, in the absence of direct fuel injection, piston motion alone. High speed camera footage (2000 fps) of a stoichiometric homogeneous charge ignition and propagation, with piston motion, was taken using a Vision Research Phantom camera. The flame kernel and subsequent propagation are shown in the four panels of Figure 5-14. Some slight distortion of the flame front is present due to heat transfer along the (horizontal) electrodes, and due to the disruptive effect of the (vertical) PSC capillary tube. However, the flame remains predominantly circular demonstrating a lack of turbulence which would wrinkle the flame front significantly.       Figure 5-14  Flame Propagation for a Stoichiometric Mixture in RCM Cylinder (time listed after spark) 5▪13 J43M INJECTOR CALIBRATION The Westport J43M injectors were used to supply the fuel for the direct injection and partially stratified injection events. The hole size for each injector was designed using the assumption of an isentropic expansion of a prefect gas through the injector orifice. Iaconis [374] first proposed this for the Westport J41 injectors (the precursor to the J43M’s) and obtained good agreement with experimental data. Further, an assumed loss of 10% in stagnation pressure through the injector (as recommended by Hill & Oulette t=0.5ms t=2.5ms t=7.5ms t=17.5ms ▪ Chapter 5 ▪ Repeatability & Calibration Tests ▪  96 [8]) was made to identify the state at the injector sack prior to ejection. The hole for the DI injector was sized to be Ø0.75mm, while that for the PSC injector was sized at Ø0.26mm.   Figure 5-15  J43M Injector Characterisation Data and Model Prediction  The discrepancies in the recorded data and the model predictions may be attributed primarily to the two key assumptions in the model, namely: isentropic expansion and 10% stagnation losses in the injector. It can be seen that the model under predicts injection fluxes for the smaller hole injector and over predicts for the larger hole. This would suggest that the losses in the injector are related to the mass flow rate directly and so by assuming ‘across the board’ losses of 10% in stagnation properties and isentropic behaviour the discrepancy increases with increased mass flow, as can be seen in Figure 5-15. The general trends are in good agreement however, and support the use of the model as a design tool to establish the required hole size for the injectors. 5▪14 CONCLUSION Chapter 5 has detailed the experimental results that allow the correction and calibration of the LIF intensity images. Camera CCD and laser beam profile corrections have been presented, as has an appraisal of the experimental signal-to-noise ratio. An evaluation of the compression stroke polytropic exponent allows the specification of the RCM cylinder state, while turbulence in the early stages of the compression can be neglected based upon inspection of the luminescent flame front. The calibration curves for the acetone seeder, the fuel injectors and fluorescent response of the acetone mixture have also been introduced.   97 Chapter 6  IMAGE CALIBRATION   6▪1 INTRODUCTION Chapter 6 uses the LIF results from a single point in the test matrix to highlight some of the correction procedures used to ensure that the collected data is corrected for known error sources. The chapter goes on to outline the difficulties in setting-up and calibrating LIF images for high speed, highly turbulent flows. The discussions highlight the need for temperature field data if LIF images are to be accurately calibrated. Further to this, arguments are presented that rule out condensation of the acetone seed as the source of the observed calibration errors. Finally, the practical implications for the current work are presented, while quantifying the effect of the uncertainties is left until Chapter  9.  ▪ Chapter 6 ▪ Image Calibration ▪  98 6▪2 IMAGE PROCESSING PROCEDURES AND PRESENTATION 6▪2▪1 GLOBAL PROCEDURE All image processing and post-processing was carried out in Matlab (R2009a), using a combination of library functions and bespoke code. The images from the iCCD camera were stored and imported to the processing code as 12bit TIF images. The images were subsequently converted to 16bit (double precision) TIF images for processing. Figure 6-1 presents the route followed in the post-processing and correction of the LIF images.  Figure 6-1 Image Post-Processing Road Map  A background reference image is taken of a homogeneous unseeded NG charge. This image is corrected for optical distortion, laser and camera CCD effects, namely: laser sheet thickness, laser sheet profile, laser power and CCD spatial variation. The series of 51 calibration images is treated similarly, with the additional step of background light removal. This is repeated for a range of seeded NG homogeneous charges at different AFR’s. These images provide the calibration reference for image intensity. Any given experimental image is corrected for laser effects, camera CCD distortion, background light collection, and if required the distortion from passing through a curved cylinder bore. Each image is stored to form an array x×y×n where the image is x×y pixels and there are a total of n images. ▪ Chapter 6 ▪ Image Calibration ▪  99 6▪3 CORRECTION PROCEDURES 6▪3▪1 EXCITATION (LASER) ASSOCIATED CORRECTIONS The collected LIF intensity, Sf, is given by [375], Eq. 6.1:   Eq. 6.1 where the quotient represents the optical yield and the remaining terms the emitted fluorescent intensity. The exponential term in Eq. 6.1 pertains to the extent that the laser excitation energy is attenuated by the gas mixture, and is expanded upon in §6▪3▪1▪1. For a fixed excitation wavelength and optical collection system, it is apparent that the signal intensity is proportional to interrogation volume, seed concentration and laser intensity.  Eq. 6.2 To remove the volume and laser intensity dependency of the signal, the original LIF images were corrected for laser power variation, laser-sheet thickness variation and laser intensity profile. Power correction was made by recording the laser power at which the image was collected. A small (invariant) subregion of the image was used to calculate a representative intensity for the image. Given that image intensity scales linearly with power, a linear regression was performed to yield the proportionality between intensity and laser power. All images were then scaled to be at the equivalent of 300mJ (the energy level requested of the laser). The laser sheet optics are known to produce a sheet which is thinnest, at ~0.8mm, on the cylinder centre-line (also the image centreline) and varies almost linearly to a thickness of ~1.45mm at the bore. With this knowledge all intensity values were scaled to be as though they were from a constant interrogation volume 0.8mm thick. The laser sheet passes through the test section of the RCM at an angle of ~8.50; correspondingly each row of the captured image was scaled according to the laser profile of Figure 5-9 (p.89), to bring the counts in line with the profile mean.  6▪3▪1▪1 BEER-LAMBERT ATTENUATION The attenuation term in Eq. 6.1 has been dropped from Eq. 6.2. As a laser passes through an absorbing medium it must lose energy to the medium. Considering now the lightsheet passing through the RCM cylinder, the laser energy is distributed across the air (primarily nitrogen and oxygen), natural gas (with its many components) and acetone. At 248nm excitation the only chemical present with an appreciable absorption cross-section is acetone. The Beer-Lambert absorption law describes the level of attenuation due to absorption according to Eq. 6.3  Eq. 6.3 where σa is the molecular absorption cross-section, 2.2×10 -20 cm2 [376] and the pathe length l. At the image time the thermodynamic state in the RCM cylinder yields an acetone concentration (based on a ▪ Chapter 6 ▪ Image Calibration ▪  100 maximum seed concentration of 1%v/v, (cf. §5▪11) of ~2.5×1017 molcules/cm3. From this and Eq. 6.3 the intensity ratio I/I0 is ~0.99 cm -1, or 1% loss in laser intensity per centimetre at maximum acetone concentration. Given the low seed concentration in the bulk charge and the short path length for the laser at higher concentrations (i.e. the jet dimension) it was deemed that correction for laser attenuation was not required. 6▪4 RESULTS Application of the image correction procedures outlined previously yields the intensity image of Figure 6-2 for the central condition of the CCD test matrix. Under these conditions there is a background charge of RAFR = 1.75 and the timing of the DI jet should cause the PSC injection to catch the central portion of the vortex ball at the head of the DI jet.   Figure 6-2 Intensity Image for Test Matrix Central Location  6▪5 DISCUSSION 6▪5▪1 IMAGE CALIBRATION DIFFICULTIES It is clear from Figure 5-11 (p.92) that pure fuel should yield a count of ~1385, while Figure 6-2 shows counts extending to the full 12bit range of the camera. By returning to the fundamental equations that govern fluorescent intensity it is possible to assess the possible causes of the discrepancy. Assuming linear excitation, the fluorescent signal intensity is given by [258] (a recast form of Eq. 6.4):  Eq. 6.4 assuming that bath-gas composition does not affect the fluorescence. In considering Eq. 6.6 laser power and wavelength may be disregarded since they are constant for any given image and laser source and ▪ Chapter 6 ▪ Image Calibration ▪  101 presupposing a single-phase flow; the observed intensity is proportional to temperature, pressure and mole fraction, Eq. 6.5:  Eq. 6.5 It is possible to assess the temperature change in an isenthalpic process through consideration of the Joule-Thomson coefficient, µ, defined by Eq. 6.6:  Eq. 6.6 which is negative for an observed temperature rise, identically zero for constant temperature and positive for a decrease in temperature. To make an assessment of the jet temperature at the final Mach disk, which may extend as far as ~15-20 orifice diameters downstream [377, 378], the Joule-Thomson relationship may be used. Assuming that no mixing takes place over the expansion fan boundary, assuming the expansion to be isenthalpic, using the Joule-Thomson coefficients reported by Budenholzer et al. [379] for methane and applying a cubic fit to data (R2=0.9995) over the range 1800-35psi, the temperature was evaluated to be ~247K. The pressures chosen for the range over which the Joule-Thomson coefficient was applied were the stagnation pressure at the injector nozzle (equal to the reservoir pressure less 10% [374] due to friction within the injector) and the cylinder pressure at the time of injection. At the calculated temperature, the jet density is ~44% higher than the calibration charge. The cylinder charge is known to have a temperature of ~350K at the image time, and Grossman et al. [307] have shown that fluorescent yield decreases by about 35% for every 100K temperature rise. Further, the increased gas density can reasonably be expected to increase the fluorescent intensity due to effects similar to those of higher pressure. The gas density present in the jet corresponds to a pressure of ~3.5bar at cylinder temperatures that, according to the data of Braeuer et al. [380], would increase the fluorescent intensity by about 40%. The density increase, which acts through increased absorption efficiency and increased acetone number density, along with higher fluorescent yields at reduced temperatures would generate a signal strength for pure fuel of ~6600counts (assuming all three effects combine linearly). It is evident that as the jet mixes downstream of the Mach disk and before the jet recovers the cylinder conditions there would still be counts exceeding the calibration range.  Despite being able to specify this ‘initial’ condition there is no appropriate analytical approach which may be applied with certainty to downstream locations that would circumvent the need to identify the thermodynamic state for any given pixel; specification of the fuel concentration requires knowledge of the temperature field. ▪ Chapter 6 ▪ Image Calibration ▪  102 6▪5▪2 TEMPERATURE FIELD SPECIFICATION The need to specify a temperature field is evident by considering, once more, Eq. 6.4 and reintroducing the wavelength dependency of the absorption cross-section and fluorescent yield. Exciting any given region of a flow with two different wavelengths will produce two different fluorescent signals. Taking the ratio of these two signals eliminates the number density from expression Eq. 6.5 to yield Eq. 6.7  Eq. 6.7 of which it is stated in [258] that “if the right-hand-side of Eq. 6.7 is a known function of temperature for the given excitation wavelength and a pressure sufficiently close to that encountered in the flow, the fluorescence ratio measurement, with a single calibration point, yields temperature data. Temperature measurement with [a simultaneous] dual-wavelength technique enables quantitative measurement of concentration in cases where temperature variation previously precluded the possibility”.  Upon first inspection it may appear that an experimental technique exists to circumvent the calibration problems in the current work. However, the technique is highly dependent upon having accurate calibration data for the fluorescent response of acetone at different temperatures and different wavelengths. The early work on the temperature response of acetone was limited in scope (excitation wavelength, bath gas, pressure etc.) and only in qualitative agreement. Similar reductions in fluorescent intensity were reported by [307, 363, 364] at increased temperatures. The only notable exception was Wolff et al. [252], who report no discernable structure to the temperature response and report figures which have a ±15% fluctuation about their mean. Thurber et al. [258, 305, 381] made significant steps in characterising the temperature response of acetone, with later enhancements made by Braeuer et al. [380] and Kearney & Reyes [203]. All the results were in generally good agreement though Kearney & Reyes suggest a slightly lower fluorescent response than Thurber et al.. Figure 6-3 shows the data of Thurber et al, Braeuer et al. and Kearney and Reyes for 248nm excitation of acetone in the bath gas noted in the legend. The data of Grossmann et al. [307] is also replicated despite being for 3-pentanone since. as the next homologous molecule to acetone and with the same C=O chromophore it can be expected to behave similarly. The results show a good level of agreement.  The data presented covers the temperature range from ambient to ~1000K, while for the current study temperatures in the DI jet can be expected to in the order of 200-350K. Response data in this range is limited to the single, low pressure, study of Bryant et al. [311], which reverses the trend seen in the data presented above. Bryant et al. report a 6% increase in signal intensity with increased temperature from 240 – 300K. This set of observations needs further work to be reconciled with the previous findings.   ▪ Chapter 6 ▪ Image Calibration ▪  103  Figure 6-3 Acetone Fluorescent Yield Variation with Temperature at 248nm Excitation,  reproduced from data [203, 258, 307, 380]  As described previously, temperature measurement can be made by utilising the difference in the fluorescent intensity of acetone when excited at different wavelengths. This was first proposed by Grossmann et al. [307] using a 308/248nm combination since the signal intensities from each wavelength diverge at increased temperature. This was built upon by Thurber et al. [258], to produce Figure 6-4. The most significant implication of Figure 6-4 for the current work is that as temperatures approach room temperature the fluorescent signal ratio (the result of Eq. 6.7) monotonically approaches unity, negating the ability to gain meaningful temperature information.   Figure 6-4  Fluorescent Signal Ratio Temperature Variation for 308/248nm and 308/266nm Pairs. Reproduced from [258]  The problem of calibrating the signal intensity from the DI jet is compounded further in that although the jet has recovered the cylinder pressure, it is not homopynic. Given that pressure dependent reactions act through interaction probabilities and third bodies, pressure can be taken as a proxy for density. Accepting that the excited, singlet, state of acetone is rapidly deactivated to the triplet manifold with almost unity ▪ Chapter 6 ▪ Image Calibration ▪  104 efficiency by inter-system crossing (ISC, cf. §3▪11▪4 and [303, 324, 382]), fluorescent intensity has been found to be independent of pressure [328] (supported by Wolff [252] in the range 1-18atm), suggesting that collisional effects upon fluorescence for the singlet state are unimportant (be they due to bath density or composition). However, more recent works by Grossmann et al. [307], Yuen et al. [304], Thurber et al. [357] and Braeuer et al. [380] all suggest significant changes in fluorescent intensity with increased pressure above ambient due to changes in the rate of ISC. The reported pressure responses are, generally, in qualitative agreement particularly for the latter works of Braeuer et al., Thurber et al. and Yuen et al. who all report a gradual increase in fluorescent intensity with increased pressure up to ~5bar, from where there is a transition to a high pressure plateau at about 15bar. However, the inconsistencies in the data make it hard to draw any significant conclusions with respect to the effects of gas density upon the current work. 6▪5▪3 POTENTIAL ACETONE CONDENSATION The arguments presented presume a single phase flow. Consideration must be given to the potential for the acetone seed to condense. Assuming the jet to be a binary, single-phase mixture of methane and acetone, the jet total pressure is, by Dalton’s Law, the sum of the partial pressures. Assuming the system to be in equilibrium, the acetone partial pressure must equal its vapour pressure. Using the ideal gas equation of state the acetone partial pressure may be expressed according to Eq. 6.8:  Eq. 6.8 The mass quotient is known to be 0.0375 (from calibration of the seeder, cf §5▪11), while the ratio of specific gas constants is known to be 485.7 (from calculations based upon the composition data in Table 5-1 and seeder calibration concentration data). The acetone vapour pressure is thus ~1.14% of the total pressure, or ~2815Pa. Acetone is known to follow well the Antoine equation [196], Eq. 6.9, which describes the vapour pressure of pure substances, viz.   Eq. 6.9 where A, B & C are constants (with values according to Table 6-1) and T temperature (in K).  Constant kPa A 6.25478 B 1216.689 C -42.875 Table 6-1 Antoine Equation Coefficients for Acetone,                                                                                   reproduced from [383]  The values in Table 6-1 yield, at the jet temperature of ~247K, a vapour pressure of ~1963Pa, which is below the partial pressure of acetone in the jet. Consequently, it is possible though unlilely that, provided ▪ Chapter 6 ▪ Image Calibration ▪  105 nucleation sites were available, condensed acetone may result in increased fluorescent intensity due to a second liquid phase. To investigate if droplet formation was taking place the DI jet was imaged with a ×12 high magnification Nivitar lens and LaVision Compact Imager CCD camera. Shadowgraph images were taken under identical cylinder conditions to those of the LIF study and illuminated using a Photonics Analysis PalFlash 500 capacitive discharge flash. A calibration grid showed each camera pixel to be ~1.5µm square, practically limiting the diameter of detectable droplets to ~5µm.  No droplets can be seen in the images from the acetone seeded injections (Figure 6-5 (b) & (d)) or the unseeded injections (Figure 6-5 (a) & (c)). Condensation is a complex process and it is possible that droplets less than ~1µm could have formed in the jet and their presence give rise to the strong signal observed. The timescales for the condensation, the role played by potential nucleation sites, the expansion fan shock structure and the thermodynamic state make it very hard to categorically rule out condensation as the source of the strong fluorescent intensity. The droplets, if present, are unlikely to be ballistic and carried into the jet far field, the region of most interest, and are also likely to have had time to evaporate as the jet increases in temperature. In the author’s opinion the thermodynamic arguments presented previously provide stronger grounds upon which to base the calibration difficulties.    (a) (b)   (c) (d) Figure 6-5 High Magnification Shadowgraph Images of Jet: (a) natural-gas jet, far-field (b) acetone seeded NG jet, far-field; (c) NG jet, near-field; (d) acetone seeded NG jet, near-field ▪ Chapter 6 ▪ Image Calibration ▪  106 6▪5▪4 THERMODYNAMIC STATE & FLOW FIELD INSIGHT FROM CFD The fluorescent intensity measured by an iCCD camera is the combination of pressure and temperature affecting the absorption cross-section and fluorescent yield as well as the thermodynamic state of the acetone (natural gas). Given the shortcomings of experimental techniques to establish the temperature field, insight was sought through computational fluid dynamics. Mr. Ehsan Faghani, a fellow student and researcher at the UBC Clean Energy Research Centre, is conducting extensive numerical research into the injection of natural gas for natural gas engines. A previous model developed by Faghani had geometries identical to those used in the current work for the DI injection, and gave the results shown in Figure 6-6.  The model uses an irregular uniform 18,000 cell axisymmetric mesh generated in Gambit® with mesh density decreasing away from the jet origin and jet centreline. A velocity inlet specified the jet inlet boundary condition, a pressure outlet downstream specified the flow exit, while symmetric boundaries along the jet centreline and faces and walls on the remaining faces closed the computational domain. The transient governing equations were solved in Flunet® with an implicitly time stepped k-ε Reynold’s stress closure of the upwind differenced Reynold’s averaged Navier-Stokes equations. Walls were treated using a ‘law of the wall’, while convergence was deemed to have been achieved when the residuals on all parameters was <10-4.   The injection pressure and chamber conditions matched those of the experimental study.  Downstream of the expansion fan the jet can be expected to behave as if it were correctly expanded. As outlined by Ouellette & Hill [384] the isentropic Mach relations of Eq. 6.02 can be used with Ma=1 to specify the thermodynamic state of the jet at the Mach disk (with gas composition data according to that of Table 5-1 & Table 5-3).    Eq. 6.10 which yield critical values at the Mach disk of: pressure 995psi (6.85MPa); temperature 262K; and velocity 407m/s. Further, Ouellette & Hill are able to show that the equivalent jet diameter, de,- that which maintains the mass flow rate and injected momentum is given by  Eq. 6.11 and it is this diameter of 4mm with the critical values listed that are specified for the jet inlet in the CFD model.  ▪ Chapter 6 ▪ Image Calibration ▪  107 It is apparent from Figure 6-6(a) that although the cylinder pressure can be assumed recovered, the thermodynamic state within the jet is highly nonuniform, which in turn yields itself in large density gradients. Temperatures at the jet origin are similar to those predicted from the Joule-Thomson calculations through ~10-15K warmer. This is significant in that the value of 260K at the jet origin when substituted into the Antoine equation, Eq. 6.9, gives an acetone vapour pressure of  ~4480Pa. This is well above the partial pressure of acetone in the jet (2815Pa), and supports the experimental indications of no acetone condensate.  Of more significance is the insight provided by the density plot. The high flow velocity means that since the pressure is uniform throughout the domain, the fluid density must adjust to match the changes in velocity. This interplay shows the fallibility of equating lower temperatures with higher density and its effects upon acetone number density and fluorescent intensity.   Figure 6-6 CFD Gas Injection Predictions of (a) Temperature, (b) Density and (c) Velocity (a) (b) (c) ▪ Chapter 6 ▪ Image Calibration ▪  108 6▪5▪5 THE ROLE OF VELOCITY IN HIGH SPEED LIF IMAGES. The jet velocity ranges from ~400m/s at the jet inlet to ~200m/s at the RCM cylinder centreline (z~0.05m), where the PSC would be anticipated to be, and where any LIF image would be of most interest. The image of Figure 6-2 is comprised of pixels which correspond to 0.25mm × 0.25mm, and which were exposed for 1000ns. It is accepted practice to circumvent the delay and jitter in triggering the laser source by exposing the camera and triggering the laser pulse to fall within the exposure duration (cf. experimental timing diagram, Figure 4-7 p.67). With an exposure of 1000ns, phosphorescence should be effectively shuttered since it has a significantly longer lifetime than fluorescence at 30µs-200µs for emission from the hot triplet and normal triplet states respectively (cf. §3▪11▪4). The laser pulse duration is ~20ns, and with rapid activation of the acetone it can be expected that any given molecule within the interrogation volume, defined by the pixel dimensions and the laser sheet thickness, will undergo only a single fluorescent event. Given the almost unity probability with which ISC takes place (fluorescent yield Φfl~0.0017 ) it is unlikely that any given molecule will fluoresce return to its ground state, be re-activated and fluoresce again, it will in all likelihood undergo ISC to the triplet manifold and relax from this state.  In the case of the DI jet (and also the PSC jet) the discussion must be framed in terms of the number of molecules which are open to excitation over the course of the laser pulse. The interrogation volume, although stationary in space, does not contain the same gas throughout the laser pulse and camera exposure because of mass transport into and out of the volume. With fluid flow there is the opportunity for more acetone molecules to be transported into the interrogation volume and excited prior to the laser pulse ending. The acetone molecules that were in the volume originally are excited to the singlet state and through ISC go on to populate the triplet manifold. Those left in the single state may fluoresce and be reactivated, however given that the majority (99.83%) of the acetone is still residing in the triplet state any acetone transported into the pixel volume will also undergo excitation, while those that exited the volume are likely to be in the non-fluorescent triplet state and represent no loss in molecules which would contribute to the fluorescent signal. The mass transport is in effect increasing the ground state population from which activation and subsequent fluorescence may take place. There now exists a larger number of excited molecules than were the excited complex to be populated from a ground state population solely comprised of original molecules. These can subsequently fluoresce and have their emissions captured during the (long) 1000ns exposure. It is these extra advected acetone molecules which may be ascribed to the stronger signal observed in the DI jet when compared to the calibration case. 6▪5▪5▪1 PIXEL VOLUME MASS EXCHANGE Mass exchange through entrainment is only part of the consideration. The regions within the jet that demonstrate excessive intensity are towards the core of the jet, away from the shear layer where ▪ Chapter 6 ▪ Image Calibration ▪  109 entrainment would be most prevalent. The entrainment coefficient varies from 0.024 to 0.118 depending upon the original jet velocity profile, though is often accepted to be 0.057 [385]. At this level, over the course of the 20ns pulse, the mass exchange in the pixel volume is negligible and the can certainly be assumed frozen on the basis of entrainment velocity. Making an assessment based on entrainment further misses the role of mixing in image calibrations, since it need not be ambient gas that is brought into the pixel-volume to increase the signal it need merely be any ground-state acetone doped gas that was not originally present.   In this way it is important to now consider the role played by the large coherent structures in the jet and the associated energy cascade. The energy held by the jet has a distribution similar to that shown in Figure 6-7, with eddy lengthscales ranging in size from the jet width (L ) to the Kolomogorov lengthscale, η. Of most concern are the eddies associated with inertial subrange.   Figure 6-7 The Turbulent Energy Cascade  The increase in signal intensity over the calibration case can be attributed not to mixing per se, but the transport of fuel (acetone) into the pixel-volume, necessitating consideration of the local flow. The eddies associated with the dissipative range, and the Kolomogorov lengthscale, are wholly within the pixel volume and cannot be responsible for transport into the pixel; yet the lower lengthscale of the inertial subrange is defined with respect to the Kolomogorov lengthscale. Given that the jet Reynold’s number is independent of downstream location it is appropriate to scale the flow against this and in doing so Pope [386] is able to show that: ▪ Chapter 6 ▪ Image Calibration ▪  110  Eq. 6.15 where  and l0 is of the order L. For the DI jet the characteristic large eddy timescale is ~20µs, and with the exposure of Figure 6-2 set to 1000ns the eddy rotation is deemed too slow to significantly affect transport into the pixel. With the majority of the turbulent kinetic energy held in the large structures and with energy destruction only on the Kolomogorov scale and below, the inertial subrange, which is responsible for the transfer of energy through vortex straining and the conservation of angular momentum [387], is responsible for the mass transport into the pixel-volume. The inertial subrange extends from ~l0/6 at the energy/inertial subrange interface (lEI) to ~60η at the inertial/dissipative interface (lDI) [386]. At these limits the characteristic lengthscales are approximately lDI~0.024mm and lEI~2mm, precisely the range over which transport into the pixel-volume is of concern. Using a scaling based on length [m] and dissipation rate [m2/s3] Pope deduces that  Eq. 6.16 which, with substitution of lDI and lEI for l, yields velocities from 55-220m/s and timescales of 395-9000ns in the inertial subrange. With interrogation volume dimensions of  0.25 × 0.25 × 0.8 mm the eddy sizes of concern are those with about twice the face dimension (0.25mm) or of the laser sheet thickness (1mm). These dimensions represent flow velocities in the range ~150-200m/s with timescales of ~3000-5000ns. As with the entrainment velocity, over the 20ns pulse of the laser, the velocities and timescales alone are not able to account for the increase in LIF intensity alone.  The out of calibration problem may be circumvented to a large extent by reducing the iCCD exposure duration, and this was used subsequently for all data collection in this study. The laser trigger delay was established by finding the threshold at which a LIF signal was captured. With this established (to be 980ns), and accepting that the jitter in the laser pulse is 10ns the minimum camera exposure was set to 40ns. This allowed the full laser pulse to be captured even at the maximum extents of the laser jitter. By reducing the camera exposure the extent to which ‘fresh’ acetone was entrained into the interrogation volume was not reduced, but the likelihood of capturing its fluorescent light was. ▪ Chapter 6 ▪ Image Calibration ▪  111 6▪6 CONCLUSIONS Having made corrections for experimental errors, and having discounted seed condensation as the source of the observed calibration errors, there are two distinct aspects at work in attempting to make quantitative concentration measurements by comparison to a calibration image. The first is that the thermodynamic state cannot be equated in the jet and calibration cases, and secondly that there is significant motion in the jet case that is not present in the calibration image. When the camera shutter speed is reduced to its experimental minimum, the jet images fall within the range of the calibration curve. This would suggest that although the thermodynamic uncertainty is still present it is within the experimental error, and that although the mass exchange and mixing arguments presented do not offer a complete explanation of gas motion is the predominant cause of the calibration difficulties. All existing work on high speed flows [158, 161, 163] use laser systems that allow much smaller interrogation volumes to be formed (typically about 50 × 50 × 200µm), so limiting the extent to which mixing needs to be considered. Furthermore, the studies do not seek concentration measurements, they are designed for the most part to offer qualitative structural information [158, 161, 163], the remainder providing temperature field [165] or velocity insights [388].  112 Chapter 7  PARTIALLY STRATIFIED CHARGE, DIRECT INJECTION & INSERT FUEL JETS   7▪1 INTRODUCTION Chapter 7 is concerned with the initial findings of the central composite test matrix. Using the example of the fuel concentration fluctuations, the shortcomings of the RCM optical set-up in imaging the full RCM bore are presented. The inability to completely correct for background noise and scattered light interference is discussed but, despite the shortcomings, some basic observations about the interaction of the PSC and DI jets are made. With the limitations identified the chapter goes on to discuss the results of the LIF images obtained for a capillary injected PSC jet into air and an insert generated jet into background charges of RAFR 1.5-2. The behaviour of these injections, without the presence of a DI fuel jet are used to form the baseline from which comparisons in behaviour will be made in the Chapter 8. ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   113 7▪2 DATA PRESENTATION As is often the case with visualisation studies, there is a significant challenge in presenting the data from across the full range of the CCD test matrix. The LIF images are presented according to the schematic shown in Figure 7-1, with laser illumination from the right-hand edge of the image.  Figure 7-1 Data Visualisation Schematic  Unless otherwise noted the data displayed is from the central composite test matrix described in §4▪8 where each data point corresponds to 51 realisations, except for the central condition that had three repeats totalling 153 realisations. In some circumstances data will be presented from the central region of the image only. In the images from the insert study the insert sits flush with the firedeck and is not evident in the images. In the case of the capillary injections an image mask has been applied to mask the images where the laser sheet was obscured by the capillary or where the capillary was located. The cylinder centreline is located at 50mm in the full bore images and at 0mm in the central detail images. 7▪3 INEFFECTUAL BACKGROUND CORRECTION Background light correction adjusts an image for light that originates from inelastically scattered laser light. The correction is made by collecting 51 images of the RCM cylinder filled only with air. Following correction for optical distortion and camera CCD variation the images are arithmetically averaged to yield a mean image such as the one shown in Figure 7-2.  ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   114  Figure 7-2 Background Correction Image  The high count along the top edge of the image is the reflection from the RCM fire-deck, while those near the centre are from the laser-sheet impinging on the PSC capillary tube. The vertical streaks are reflections from the internal bore of the quartz window. The common mindset within the laser diagnostics community is that subtraction of this image from subsequent images will provide an image that is devoid of scatted light and background interference. Application of this procedure to Figure 7-3 yields Figure 7-4, from which it is evident that the background correction is failing to completely remove background interference. This happens because the DI jet produces fluorescent light that enhances the reflections of Figure 7-2. In this way, the elastically scattered light will always have more spatial variability and higher levels than the background image can correct for.   Figure 7-3 Raw LIF Image Counts [-] Counts [-] Firedeck Reflections Capillary Tube Reflections Cylinder Bore Reflections ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   115  Figure 7-4 Ineffectual Background Corrected Image 7▪3▪1 ROLLING BALL SMOOTHING In optical investigations of liquid fuel sprays, it is common practice to use a rolling ball smoothing technique to smooth the image backgrounds, a technique itself borrowed from biomedical research and diagnosis. The technique involves the application of an image ‘erosion’ (removal of high count pixels) followed by a dilation (smoothing) to remove the background interference. The erosion is applied by specifying a ‘structuring element’ which is a local mask describing how the image pixels are to be altered within the mask. The processing attempted in the current study used a ball element, an element which is circular in the xy plane and which maps the pixel values within its radius to the surface of the ball. The element is traversed across the image (hence the term rolling ball) in both the x and y directions and any pixel counts which fall outside the ‘radius’ of the ball are brought down to the value of the ball radius. This is more readily seen in Figure 7-5 where the ball structuring element will accurately follow the larger features of the image but ‘erode’ the sharp peak. The dilation process involves ascribing the same value to all the pixels within the structuring element, based on a number of possible rules. In doing this the pixel counts are spatially smoothed to reduce image gradients.  Figure 7-5 Rolling Ball Erosion,  reproduced from [389] Counts [-] ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   116 These techniques, although very powerful, could not provide any significant improvement in the image quality for the current work. This may be attributed to the fact that the noise often changes gradually, limiting the effectiveness of the erosion. Of greater concern is the effect of the rolling ball upon the structure and gradients at the jet boundary- the very aspects of the jet structure that this study is concerned with. Application of the rolling ball to the jet would destroy the structures under investigation and as such an image mask would need to be applied so that different processing procedures could be applied to different parts of the image. The difficulty in doing this is identifying the pixel threshold value at which the edge of the jet is found and the background begins. This is further complicated by the occurrence of the threshold values in the background noise facets. No acceptable method was found to circumvent these interference concerns. As a result, the region of interest was reduced so that, for the most part, LIF images were collected from the portion of the cylinder not affected by background interference. This also afforded a higher spatial resolution that provided better structural clarity and helped to reduce the adverse affects of pixel mass exchange addressed in Chapter 6. 7▪4 PARTIAL STRATIFICATION WITH DIRECT INJECTION- FULL BORE IMAGING 7▪4▪1 CALIBRATION The investigations into the SNR effects of different on-chip binning approaches (§5▪10) gave the two calibration lines of Figure 7-6 and were constructed using an approach identical to that outlined in §5▪9. It is apparent that a 4×4 bin provides a stronger signal and offers a more favourable SNR. To increase image signal strength a 4×4 bin was used in all studies despite the loss in spatial resolution.   ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   117 Figure 7-6 4×4 and 2×2 On-Chip Binning Calibration Lines, Inset Detail for Low Signal Intensity  The 4×4 calibration line takes the form of Eq7.1  Eq. 7.1 where Nfuel/Ntotal is the normalised fuel number count and Sf the signal intensity. The inset in Figure 7-6 shows that the calibration may be less accurate at lower concentrations, though is still within the experimental error and appears as a ‘y-intercept’ of ~56, not the camera noise floor of 87.  7▪4▪2 CONCENTRATION FLUCTUATIONS The ability to conduct detailed analysis in the region of the DI-PSC interaction is limited because of the poor spatial resolution that results form a 4×4 on-chip bin (pixel dimension 0.5×0.5mm), while inference of the overall structure of the jet and its effects upon the cylinder charge are limited by background interference. The most informative images to consider are those of the normalized root mean square fluctuations in RAFR. Detailed methane fuel jet structures are reported by Bruneaux [390], as are the r.m.s fluctuations. The images of Figure 7-7 show, for the most part, the PSC jet interacting with the bulk of the head vortex or its trailing edge. The late DI injection (Figure 7-7 t = -5CAD) shows that the PSC jet is not catching the head vortex along its leading edge but about 8mm after it.   ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   118   Figure 7-7 Normalized r.m.s Fluctuations in RAFR for Direct Injection with PSC at Relative Time Indicated t = -5 CAD t = -3.53 CAD t = 0 CAD t = 3.53 CAD t = 5 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   119  The jet shows fluctuations in RAFR that are very similar in position and magnitude to those reported by Bruneaux [390]. The leading edge of the vortex ball shows fluctuations on the same scale as the concentration value.  The fluctuations around the core of the DI jet are larger than reported by Bruneaux but less significant to the current work, since the DI-PSC interaction is primarily driven by the shear layer at the boundary of the jets. The inner core of the DI jet shows fluctuation levels of almost zero. This contradicts previous works (cf. §2▪2▪1) which report normalised concentration r.m.s. levels on the jet centreline of about 20%. The current zero levels point to the calibration difficulties discussed in Chapter 6. The thermodynamic state of the jet and potential mixing can produce LIF intensities higher than that allowed by the calibration line of Figure 7-6. In these cases the calibration procedure artificially brings the count down to the maximum allowable, that corresponding to pure fuel. This process removes the ability to assess the fluctuations in the jet core and may contribute to the high fluctuation levels seen around the invariant core.  The interdiction of the PSC jet along the top edge of the DI jet perturbs the DI jet along this edge. The PSC injection itself is seen to have normalised r.m.s fluctuations in RAFR in the range 0.8-1.0 in all cases but there are differences in how the fluctuations are manifested in the DI jet shear layer. At the latest timing (t = -5 CAD) the PSC injection is seen to increase the fluctuations along the top edge of the DI jet, almost extending to the location on the leading edge where there are significant r.m.s. fluctuations. These fluctuations may be attributed to the increased velocity at the edge of the vortex ball as fluid circulates back (cf. schematic of jet structure Figure 2-1, p.10), entraining the capillary delivered fuel. Earlier injections, at t = -3.52 & 0 CAD shows the capillary delivered fuel causing little perceptible difference in RAFR fluctuations from those which may be expected in the DI jet alone. These timings coincide with the trailing edge of the vortex ball, at which point the flow velocities are more closely aligned with the PSC injection axis and as a result cause less shear and lower observed fluctuations, although some shearing would still be expected. As the DI injection timing is advanced still further (t = 3.52 CAD) and the PSC delivered fuel intersects the DI jet at the interface of the head vortex and the trailing jet, the flow once more starts to shear the PSC jet, increasing the observed normalized RAFR fluctuations to ~0.6-0.7. At this location the PSC injected fuel is serving to increase the local entrainment into the jet and then into the vortex ball since it is known that the vortex ball is fed from the trailing jet, although to what extent is still unclear because of the entrainment implied by the fluctuations on the leading edge. It is apparent that, as the DI jet develops, the level of normalized RAFR fluctuation decreases, though without reference to the absolute levels, it is not apparent to what extent leading edge entrainment is diminishing. This is not the focus of the current work and comparison becomes difficult because of the background noise facet that cuts through the leading edge of the jet for the earliest two timings. In the earliest timing, where the PSC jet intersects the ‘steady state’ portion of the DI jet (according to Turner’s model, §2▪2▪1), the fluctuations ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   120 in normalised RAFR are shown to drop once more as shearing of the PSC jet is now derived from the jet velocity alone with no role played by the vortex ball. The nature of the perturbations generated by the PSC jet is likely governed by the local flow pattern at the location of the PSC-DI interaction, the details of which would have to be sought from laser velocimetry techniques such as PIV of LDV. 7▪5 PARTIAL STRATIFICATION WITH CAPILLARY TUBE INJECTION 7▪5▪1 AIR FUEL RATIO The images of Figure 7-8 show the PSC jet generated by fuel injection from a capillary into the RCM chamber when filled with air. The timings are comparable to those used in the central composite test matrix. The RAFR images are generated from the calibration line of Figure 7-6 with 4×4 on-chip binning and have pixel dimensions that correspond to 0.0625×0.0625mm. The capillary exit is located at ~0.5mm from the image (cylinder) centreline and ~2.5mm off the firedeck (image top edge). The high count levels to the right of the image mask and the ‘spike’ on the bottom right corner of the mask are attributable to light scattered from the PSC capillary and cylinder bore. These facets cannot be fully corrected for.  The most notable feature from the set of images in Figure 7-8 is that the jet appears to have parallel edges and shows little sign of increasing width with downstream position. This would imply little entrainment and mixing. For the latter three timings, the thickness of the layer containing a fuel concentration gradient is almost constant and on the order of 0.25mm thick. In all cases it is apparent that the fuel undergoes most mixing along the leading edge of the jet and is starting to generate a stoichiometric mixture at about 7mm from the firedeck. The fuel-rich core of the impulsively started jet persists for longer than may be expected and may point to plug flow which could also explain the slow growth in jet width. The colour- scale appended to the image is also likely to be masking some of the jet structure. The background air charge should be displayed with an infinite RAFR but for image presentation an upper limit of 2.5 was imposed so that the jet structure could be presented, though this is likely to be obscuring the leaner portions of the jet. This difficulty can, to a certain extent, be circumvented by representing mixture concentration by mixture fraction which, by definition, covers the range zero to one. However, with this approach, a lot of the detail pertinent to combustible mixtures is compressed into a small range of values and when interpreting data relevant to engines, air-fuel ratio or equivalence ratio are more widely understood and interpreted. Data presentation will be made using the measure that best supports the argument being made.  ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   121   Figure 7-8 RAFR for PSC Jet from Capillary Tube at Times Indicated  It has been demonstrated that signal strength is proportional to fuel concentration and, in considering the signal intensity insight to the PSC jet concentration structure is obtained. Figure 7-9 shows, for the four times indicated, the fluorescent signal intensity from the centreline of the PSC jet against the axial position from the fire deck. The profiles are not those which may be expected for an impulsively started jet or the initial transient stages of an injection. t = 0 CAD t = 3.52 CAD t = 7.06 CAD t = 10.62 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   122   Figure 7-9 PSC Capillary Injected Jet Centreline Count Levels (Broken Lines Indicate 95% Confidence)  The intensity (concentration) decay with axial position is not as described by previous authors (cf. §2▪2▪1). The profile at t=0 CAD shows a decay which is similar to that expected form Eq. 2.3 or Eq. 2.4 from ~3.5mm onwards, yet upstream of this location there is an almost constant count level. The count is not that which corresponds to pure fuel and may be lower due to the bulk air charge being forced into the PSC capillary as the piston compresses the cylinder charge prior to injection. At t=3.52 CAD the count is seen to increase dramatically downstream of the PSC capillary exit before starting to decay, a decay that appears to be almost linear and not the expected hyperbola. The feature could also result from inaccurate background light correction. This structure persist as the flow penetrates further, although the maximum count level decreases and the profile broadens. These observations may be attributed to two different causes. The first is that the flow from the PSC capillary may be that of a plug. Ahead of the plug is gas that was forced into the capillary prior to injection and has undergone some mixing with the fuel already in the capillary to produce the mixture of the concentration observed in the first timing. As the plug, which was upstream of the mixed gas and which has a higher fuel concentration, emerges from the capillary exit it starts to mix with the cylinder air charge. As the plug continues to travel into the cylinder it mixes further such that its concentration reduces, since it is not ‘fed’ from behind with fuel to maintain high concentrations. The observed intensity variation may also be explained by the variability of the injection t = 7.06 CAD t = 10.62 CAD t = 0 CAD t = 3.52 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   123 events. The solenoid used to control the PSC injection is known to be variable in its response time and its opening profile. These two factors will deliver jets of different structures and at different times. This is especially true of the early stages of the transient before differences in jet structure have been lessened by mixing.  The penetration behaviour of transient jets has been described by Hill & Ouellette [8], and comparison with their results can provide further insight. Hill & Ouellette provide scaling arguments which show the jet penetration, Zt, to be proportional to the momentum injected and time elapsed, with the constant of proportionality, Γ, termed the penetration constant and equal to the pure constant 3.0±0.1 [8], viz.:  Eq. 7.2  It is apparent from Eq. 7.2 that a plot of  vs. t½ will yield the jet penetration constant. Based upon the penetration values from Figure 7-9 (i.e. the location at which the intensity drops to background) and the approach of Ouellette & Hill [384] to calculate the momentum injection rate, the jet penetration constant was evaluated to be 3.87. Although the linear fit has a regression coefficient of 0.9212, the data may still not be linear. The evaluated penetration constant is close to Hill & Ouellette’s value and may suggest jet-like behaviour. However, a lack of literature on emergent plugs does not allow comparison of characteristic features like spread rate, concentration profile and penetration.  Figure 7-10 Capillary Injected PSC Jet Penetration; (•) PSC Data (-) Least Squares Fit  There is insufficient data in the current study to truly establish which of the two effects suggested is causing the discrepancy from accepted jet structure or if it is a combination of the two. The arguments 1  3. 87   ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   124 supporting the repeatability influences are built upon further through consideration of the jet concentration fluctuations. 7▪5▪2 CONCENTRATION FLUCTUATIONS When considering the interaction of the PSC with the DI jet the levels of concentration fluctuation will be important and, as such, a baseline is presented here for the capillary injected PSC jet. The concentration fluctuations extend to about 7mm from the firedeck at t=0 CAD. This is about 2mm further into the chamber than the concentration image of Figure 7-8 at t=0 CAD would suggest and points to the variability not in the jet itself but in the ability of the solenoid to trigger the jet repeatabley. The jet displays fluctuations which are not wholly consistent with those seen for the DI jet in §7▪4 or those reported elsewhere [390]. At time t=3.52 CAD a fluctuating ‘core’ and large fluctuations on the leading edge of the jet can be seen, as would be expected from the work of [390]. However, in the two earliest cases, the area over which the fluctuations extend is not confined to the jet leading edge because of the variability in jet penetration for a set injection time. At the later timings (t=7.06 CAD & 10.62 CAD), the fluctuations appear to have a spatial distribution much more akin to those of the concentration field.   This is, again, likely due to the inability of the PSC solenoid to deliver a repeatable jet and the subsequent spatial variations for each realisation over-riding any jet structures. There is in both latter cases a slight suggestion of an invariant core with normalised RAFR r.m.s levels of about 0.25 lower than the region surrounding it, though to assess this in detail would require further investigation.  The observed fluctuation structure would not support the plug flow argument presented earlier, since it would not be expected that the plug show concentration fluctuations increases towards its centre; the fluctuations should be confined to the shear layer around the boundary of the plug [391]. These concerns, and those raised in §7▪5▪1, cannot be fully explained, though for the current study provide limited impediment since it is the interaction of the PSC jet, whatever form it may take, with the DI jet that is of concern, not the jet structure itself.  ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   125   Figure 7-11 Normalized r.m.s Fluctuations in RAFR for Capillary Injected PSC at Times Indicated 7▪5▪3 CONCENTRATION GRADIENTS It has been posited in previous work on the combination of PSC with DI that excessive fuel gradients limit the ability of the PSC charge to ignite [7]. The PSC fuel concentration gradients (as RAFR/mm) generated from a capillary injection are displayed in Figure 7-12. The images show that with increased time the concentration gradients reduce by ~50% (from t=3.52 CAD to t=10.62 CAD) since there is increased time for mixing to dilute the injected fuel; however, the reduction in the observed gradients from t=3.52 CAD to t=10.62 CAD appears excessive. The electrodes to ignite the PSC charge can be expected to be on the cylinder centreline, about 4mm from the firedeck and at this location the concentration gradients are in the range 1-3 λ/mm depending upon the spark timing. This indicates the necessity to accurately control the t = 7.06 CAD t = 10.62 CAD t = 3.52 CAD t = 0 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   126 location of the spark within the PSC jet if the mixture is to be within the fuel flammability limits but also the need to ensure that jet delivery is repeatable.  To circumvent the extent to which these spatial constraints must be imposed, a PSC ‘insert’ has been designed to increase the mixing between the PSC jet and the cylinder charge. The jet produced by the insert is the focus of §7▪6.    Figure 7-12 Spatial Fuel Gradients for Capillary Injected PSC at Times Indicated t = 10.62 CAD t = 7.06 CAD t = 0 CAD t = 3.52 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   127 7▪6 PARTIAL STRATIFICATION WITH SPARK PLUG INSERT INJECTION 7▪6▪1 JET DEVELOPMENT  Figure 7-13 Jet Development for Insert Injected PSC at Times Indicated  The images of Figure 7-13 point to some design limitations of the PSC insert. The insert is located by a thread and as such the angular orientation of the insert is not controlled. The insert is tightened to ensure a gas-tight seal and to align the face of the insert with the RCM firedeck. With these two constraints there is no way to ensure that one of the eight gas flow paths aligns with the cylinder centre-line and laser sheet. The images of Figure 7-13 show one gas jet to enter the laser sheet at about -3mm and a second on the cylinder centreline (not apparent at t=3.52 CAD).  It is immediately apparent that the concentration field is not that of a single jet or the multiple jets that the insert would be expected to produce; the fuel concentration profile is much wider than for a jet. For any given axial location the jet width is larger than would be expected from the spread rates reported in the literature (cf §.2▪2▪1). To a certain extent this is to be expected since the individual jets formed by the insert mix prior to entering the cylinder and, as such, there exists the potential for any of the jets to undergo entrainment from other jets or the small volume that forms the ‘chamber’ of the insert. Considering only the jet with origin at -3mm, it can be seen that as time increases the fuel concentration reduces and the width takes on a form which is closer to that expected of a jet. These observations point to the likelihood of the jet transecting the laser sheet at an angle rather than passing along the sheet (as is the case with the capillary injected PSC). At t=3.52 CAD the structure may correspond to that of the vortex ball at the head of the jet and as the ball passes a structure more like that of a steady jet is observed. This hypothesis is not without contradiction from the images however. The drop in fuel concentration would not be explained were it solely the passing vortex ball and the subsequent jet that were being visualised. t = 3.52 CAD t = 7.06 CAD t = 10.62 CAD ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   128 For similar reasons to those presented for the capillary injected PSC it is possible that some type of plug flow is present or, given the mixing generated by the insert, some combination of plug and jet flow. Furthermore, the injection for the insert is controlled by the same solenoid as the capillary injections and the same concerns about the injection repeatability persist.  The images of Figure 7-13 also demonstrate a number of other features that are of concern to the insert design and its efficacy. The second jet, which is apparent on the centreline, at no stage shows an origin on the cylinder firedeck (image top edge), nor does it appear in the earliest timing (t=3.52 CAD) and yet it shows greater penetration than the jet at -3mm for the latter two timings. The lack of contact with the firedeck and larger penetration suggest that the jet is coming from further out of plane than the jet at         -3mm and as a result has undergone more mixing to bring its fuel concentrations down from the higher levels observed in the jet at -3mm. However, the jet clearly demonstrates the large spatial extent to which a stoichiometric (or close to) mixture has been developed when compared to the spatially limited jet from the capillary injection. There is a clear spatial distinction between the two jets and this is preserved downstream. This suggests that the desired mixing and coalescence of the individual jets from the PSC insert is not taking place and the orientation of the holes in the insert needs further design consideration.  Of further import is the lack of evidence of a jet to mirror that at -3mm. There is no jet seen at any location on the right-hand side of the image, even as far downstream as 18mm. Investigation of the insert’s alignment with the laser sheet and firedeck showed the capillary that delivers fuel to the toroidal annulus which supplies the individual injection points (cf. Figure 4-4, p.64), was closest to the jet at - 3mm. Given the high mixture strength of this jet, the relative weakness of the jet on the centreline and the lack of a jet at ~3mm, it can be surmised that the flow within the insert is significantly affecting each jet produced by the insert. It could also be possible that the orifices furthest from the delivery capillary are receiving such little fuel that they are not effectively forming jets. 7▪6▪2 SPATIAL FUEL GRADIENTS Gorby [6] was able to ignite the PSC charge more repeatable with a weak background charge of RAFR = 1.79. The images of Figure 7-14 present the PSC injection from the insert at a constant timing of t=7.06 CAD into bulk charges with the fuel concentrations noted. Comparison with Figure 7-12 shows the concentration gradients to be about an order of magnitude lower with the insert, although their overall structure is comparable to that of the capillary injected PSC jet (notwithstanding the previous assertions on insert jet structure). The jet at -3mm appears to show very similar concentration gradients no matter what the background charge and may point to the fact that the shear layer, where the background charge would be expected to have an effect on fuel gradients, is particularly small for the insert jets at this location. The jet on the centreline is seen to have a decrease in fuel gradients for λbulk=1.75 & 2 when compared to the ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   129 case with λbulk=1.5. This may, at first, seem contradictory since it would be expected that a larger difference between the jet and bulk fuel concentrations exists. However, the observed result may point to the effect of entraining into the jet a mixture that is significantly different in fuel concentration and, as such, serves to reduce the jet concentration, bringing it closer to that of the bulk charge, thereby reducing the observed gradients. This may be expanded upon further by considering the relative-fuel-air distribution of the jets in different background charge concentrations.   Figure 7-14 Fuel Concentration Gradients for Insert Injected PSC With Bulk Charge RAFRs Indicated  7▪6▪3 BACKGROUND CHARGE EFFECTS  Figure 7-15 Background Charge Effect upon Insert Jet Mixture Distribution  λbulk = 1.5 λbulk = 1.75 λbulk = 2 λbulk = 1.5 λbulk = 1.75 λbulk = 2 ▪ Chapter 7 ▪ Partially Stratified Charge, Direct Injection & Insert Fuel Jets ▪   130 The assertion that entrainment of a weak mixture serves not to enhance gradients but encourage the ‘mixing out’ of an otherwise rich mixture to that of a stoichiometric mixture would appear to be borne out by Figure 7-15. The images of Figure 7-15 depict the same injection conditions as those of Figure 7-14 but now present the relative air-fuel ratio. As the background charge concentration is increased from λbulk=2 to λbulk=1.5 the spatial extent of the stoichiometric contour is seen to increase. This implies that the background charge concentration needs to be matched to the PSC fuel concentration profile to ensure that the jet is sufficiently diluted, but not overmixed. With no explicit control over the range of background charges tested by Gorby the improvement in ignitability of the PSC with a weak background charge may be attributed to the formation, through mixing, of a larger region of roughly stoichiometric conditions, particularly when the stochastic nature of jet structures is considered. 7▪7 CONCLUSION Despite the stochastic nature of jets, a high level discussion of the DI jet behaviour was made with respect to its normalised concentration fluctuations, providing a basic understanding of the DI-PSC interactions and, in particular, the extent to which fluctuations vary depending on where the PSC jet intersects the DI jet. The findings in this Chapter serve as a basis upon which the interaction between the DI jet and PSC may be assessed. The limited ability of the RCM optical set-up to provide data pertaining to the full cylinder bore has been demonstrated, as has the difficulty in correcting the background light interference. It has been shown that the PSC insert is not generating the mixture structure that would be expected from its design and the lack of coalescence of the jets necessitates further consideration; particularly with regard to how the mixture is ignited and how the kernel propagates into the homogeneous mixture or from one PSC jet to another.    131 Chapter 8  THE INTERACTION OF PARTIALLY STRATIFIED & DIRECT INJECTION FUEL JETS   8▪1 INTRODUCTION Chapter 8 considers, with greater spatial resolution than Chapter 7, the interaction of the PSC and DI fuel jets. The mixture distribution generated by a capillary injected PSC jet and a specifically designed PSC ‘insert’ are discussed and their differences highlighted. For both PSC injection approaches assessments are made as to the effects of relative DI injection timing and bulk charge fuel concentration upon the mixture formation, fuel concentration gradients, r.m.s fluctuations in RAFR and fuel concentration by volume. At a number of locations within the interaction region the minimum, mean and maximum fuel concentrations are considered and the probability density functions for the location presented. The local scalar dissipation rate is also calculated for all nine test conditions. The implications for the potential to ignite the PSC injected fuel are considered, and comparison made between the two injection approaches to assess if the PSC insert is having the desired effect of improved mixing and ignition potential.    ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 132 8▪2 BACKGROUND The findings of the previous chapters highlighted the need to focus on the region of the PSC-DI interaction. The same central composite test matrix as outlined in Chapter 5 was repeated for injections from a capillary and PSC insert. Figure 8-1 is provided as a reminder of the nine conditions tested, where the background charges were chosen to span those of the work by Gorby and the timings, such that the PSC injection would catch the leading edge of the DI jet, the body of the vortex ball and the steady-state jet behind the vortex ball.  Figure 8-1 Central Composite Test Matrix  The data from the test matrix is presented in 3×3 arrays of visualisations, where each visualisation corresponds to its position within the test matrix, although for compactness of presentation the two data- points at t=3.52CAD are presented in line with that of t=5CAD and similarly, the points at t=-3.52 are presented in-line with t=-5CAD. For the same reason, these four data-points are presented in line with the maximum or minimum bulk mixture fraction as appropriate. For clarity these conditions are noted in each image for the RAFR visualisations presented and is implied for all other visualisation arrays unless otherwise noted. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 133 8▪3 MIXTURE DISTRIBUTION 8▪3▪1 CAPILLARY INJECTED PARTIALLY STRATIFIED CHARGE The most obvious feature of the PSC DI interaction in Figure 8-2 is that the PSC jet structure is preserved and penetrates that of the DI jet. The fluid motion of the DI jet is not enough to significantly alter the air- fuel ratio of the PSC jet. In all nine cases it is the interaction of the fuel rich core from each jet that dominates the mixture distribution in the vicinity of a potential spark location. Figure 8-3 shows the image intensity levels from the centreline of the PSC capillary injection with a DI injection. The plots are similar to those of the PSC injection alone (Figure 7-9, p.122), in that there is an observed rise to an intensity (concentration) maximum before a concentration decay occurs. In all cases a stoichiometric mixture is not reached until the far side of the DI jet, at about 9mm from the firedeck, regardless of background charge or injection timing. However, the structure of the counts differs from the PSC only case in that the concentration ‘plateau’ of Figure 7-9 (p.122) is not seen and the decay in concentration is closer to that which may be expected of a jet. In the three cases with richer bulk mixtures in the range λ=1.5-1.57 (right-hand column of Figure 8-2) the decay appears to be closer to the expected hyperbola, than in the other cases, which appear more linear. This is most likely attributable to the fact that the locus along which the points were taken is neither axially along the di jet nor transverse to it, so it could easily be expected that the profile demonstrate a shape which is the superposition of a hyperbolic axial decay and the more Gaussian radial profile.  There is no significant increase in area of the RAFR=1 contour in any of the nine cases, either due to relative timing or bulk charge concentration. The indication that timing has had no strong effect suggests that in order to maximize the area of the stoichiometric contour the leading edge of the PSC and DI jets must interact. In the case of the capillary injected PSC this is difficult because of the limited spatial extent of the stoichiometric mixture contour (cf Figure 7-8, p.121.) and that the region with an approximate stoichiometric composition is dominated by the DI jet. Considering the cases with the weakest bulk mixtures, i.e. those from λ=1.75-2 (two right most columns of Figure 8-2), there is little difference in structure along the edges of the DI jet; for example the thickness of the λ=1.5 contour (yellow) is constant at about 2mm in all cases. This suggests that, on average, mixing is confined to shear layers of similar dimensions in all cases and that the background charge strength does not largely change the spatial extent of the mixing in the DI jet. The head of the DI jet in these same cases shows some difference however, suggesting that mixing over the leading edge of the jet is more variable. In the cases with a stronger background charge (central column) the head of the DI jet appears to mix towards a stoichiometric composition over a larger area than in the case of the weaker mixtures (right column), which appear to preserve the DI head structure. This indicates that it is important to match the background charge ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 134 composition with the amount of mixing anticipated, so that the mixture composition can be optimized for engine relevant conditions.     Figure 8-2 RAFR Distribution with Capillary Injected PSC with Direction Injection t=0CAD λbulk=1.5 t=0CAD λbulk=1.75 t=0CAD λbulk=2 t=-5CAD λbulk=1.75 t=-3.52CAD λbulk=1.93 t=5CAD λbulk=1.75 t=3.52CAD λbulk=1.57 t=3.52CAD λbulk=1.93 t=-3.52CAD λbulk=1.57 ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 135      Figure 8-3 Capillary Injected PSC Jet Centreline Image Intensity 8▪3▪2 INSERT INJECTED PARTIALLY STRATIFIED CHARGE It is apparent from comparison of Figure 8-4 to Figure 8-2 that the PSC insert significantly alters the form of the interaction between the two jets. The lower momentum of the insert injected fuel, compared to the capillary injected jet, means the jet structure is more readily perturbed by the momentum of the DI jet. The distinct jet structures of the insert have now been replaced with a single region of similar stoichiometry, the form of which is affected by DI jet timing and bulk charge. For bulk charges with RAFR 1.5-1.57 (left column of Figure 8-4) a rich core persists for the injected PSC fuel. At the most retarded timing for these three conditions (t=-3.5CAD), the fuel mixture is more homogenous and has a higher RAFR, which may be attributable to the higher DI jet turbulence at this timing causing more mixing of the background charge.  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 136    Figure 8-4 RAFR Distribution with Insert Injected PSC with Direction Injection   At the most retarded DI injection timing (Figure 8-4 top row, centre) the PSC injected fuel still maintains some of its original structure when compared to all other timings; this shows the PSC injected fuel to have ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 137 been entrained into the jet, eliminating the DI jet head/vortex ball structure. In as much as the area covered by a stoichiometric mixture is larger with the PSC insert, the mixture is less rich than the capillary injected mixture and as such its combustion may be less robust. The essential role of the PSC fuel is to provide a strong initial combustion to inflame the bulk charge or, in the current case, the DI injected fuel. A less vigorous burn of the PSC fuel may leave it more susceptible to extinction from bulk charge motion or the motion imparted by the DI jet. In some individual realisations the DI jet is shown to deflect the capillary injected PSC jet, Figure 8-5. Under these circumstances the DI jet is not causing mixing within the PSC jet since it is straining the jet in a direction primarily normal to the PSC jet flow, and as such there is limited shear to induce mixing. No such deflections were observed for the insert injected PSC, since the flow is already likely to be chaotic enough that either the phenomenon would not be observed or is absent altogether.   Figure 8-5 Deflection of Capillary Inject PSC Jet  8▪4 CAPILLARY INJECTED PSC IGNITION & INFLAMMATION POTENTIAL Gorby reported difficulty in igniting the PSC jet [7] unless a weak background charge was present. It was postulated in his work that this was because the fuel concentration gradients were too large but, as was suggested in the previous chapter it may, in fact, be a matter of the background charge altering the area over which an ignitable mixture exists, particularly when stochastic variations in the mixture distribution are taken into account. Figure 8-6 shows the spatial gradients in RAFR (as mm-1). Interpretation of the images is hampered somewhat by the background reflections in the top right corner and left-hand edge of each image. Despite this, some clear trends exist. The core of the DI and capillary injected PSC jets show fuel gradients that are similar and of the order 1mm-1. The shear layer that forms the boundaries of the PSC & DI jets have gradients which are about twice this level. The shear layer of the DI jet (shown as yellow to red tones) is an order of magnitude thicker than that of the PSC jet, and increases with weakening background charge concentration. The fuel RAFR gradients in the range 2-5mm-1 occur in a region of approximately the same thickness for the leading edge and shear layers of the DI jet (~2mm), yet show no clear structure in the range ~3.5-5.5 mm-1 where they form an ensemble of levels over the area that they ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 138 cover. Furthermore, the gradients do not appear to be changed by the relative timing of the PSC event to that of the DI injection (differences along columns of Figure 8-7). In the work of Gorby the spark would have been at the core of the PSC jet, although delivered from a modified sparkplug, not a capillary. This precludes direct comparison although it would not seem, from the findings of the current work, that fuel gradients were the sole cause of difficult ignition; ignition potential is also likely to be affected local mixture variability.  Figure 8-7 shows the RAFR fluctuations normalised against the local mean RAFR. The central condition of the test matrix shows substantially higher levels of normalised fluctuation suggesting that the other conditions have not reached an experimentally converged condition; however, since it is a comparison and not an absolute value which is sought, some conclusions may still be drawn. The structure of the fluctuations show a maximum at the core of the jets decreasing towards their boundaries over which distance there is almost an order of magnitude difference. With a decrease in bulk fuel concentration the normalised fluctuations are seen to increase in magnitude (eg. top row of Figure 8-7 which has values λbulk=1.57, 1.75 & 1.93 from left to right), supporting the assertion that it is fuel concentration variability which is reduced by the background charge, in combination with the fuel concentration gradients.   The interaction of the DI and PSC jets has normalised fluctuation levels which are two to three times lower than the capillary injected PSC alone (cf. Figure 7-11, p.125), and comparable to the leading edge of the DI jet (cf. Figure 7-7, p.118). The literature on jets largely reports time averaged profiles, to which the current work shows some level of agreement. However, the work of Schefer et al. [22] shows the difference between time averaged and instantaneous profiles to be marked. Instantaneous radial velocity and concentration profiles may essentially be ‘top hat’ [392] and, as such, the average profiles misrepresent large scale coherent ‘engulfment’ rather than entrainment. Two models for the engulfment have been proposed: the axisymmetric mode [393], and the helical mode [394]. In both modes the structures are large scale, quasiperiodic, organized and exists equally [22]. The instantaneous radial concentration profile is gradually eroded by engulfment such that the ‘top hat’ profile becomes more triangular downstream of the jet origin. The concentration decay for the instantaneous profiles can be expected to show clear step-wise decreases, with the formation of ‘ramps’ and ‘plateaus’, associated with the engulfed vortices, on the scale of the local jet diameter [22].   ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 139      Figure 8-6 Capillary Inject PSC & DI Jet RAFR Gradients ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 140  Figure 8-7 Capillary Injected PSC and DI Jet Normalised RAFR Fluctuations   ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 141  The work of Schefer et al. [22] shows a demonstrable monotonic tendency towards the widely accepted mean and r.m.s. axial and radial concentration profiles, and shows the need to ensure a fully converged experimental average, as well as the stochastic nature of each experimental realisation. To elucidate this stochastic nature at a given location it is pertinent to consider the probability density function (pdf) of the relative air-fuel ratio. This was done for four different stations within the RCM cylinder as shown in Figure 8-8 and detailed in Table 8-1. Each station corresponded to an area of 1mm2.  Figure 8-8 Sample Station Locations for Capillary Injected PSC & DI Jets   Image Position (mm) Station # x y 1 -0.5 4 2 -1 5 3 -0.5 9 4 2 8 Table 8-1 Capillary injected PSC Sample Station Locations  At each station the probability density function was calculated using all 13,000 or 39,000 pixels that correspond to the 1mm2 area for the 51 or 153 images that form the data point. The full set of pdfs for all four stations over all nine test conditions is presented in Appendix C.  The stations were chosen to match locations of particular interest: 1) where a spark might reasonably be designed to be, 2) the upstream side of the PSC-DI interaction, 3) the head-vortex of the DI jet having passed through the PSC jet and 4) the shear layer in the DI jet ahead of the PSC jet. These locations were chosen to help guide future decisions about the structure of the jets PSC injection should be generating. With the exceptions noted below, the form of the pdfs for any given station were independent of injection timing and background charge. Those ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 142 presented in Figure 8-9 are taken from the central condition (λbulk=1.75, t=0CAD) of the test matrix. For stations one and two (Figure 8-9 (a) & (b)) it is apparent that there is a much more significant probability of having a rich mixture than a stoichiometric or lean one, and even within the core of the PSC jet, at station one, there is a small but finite chance of the bulk charge being present. For the PSC jet core the probability of encountering the bulk charge composition is comparable to that of encountering a stoichiometric mixture, while the probability of having a rich mixture increases rapidly for RAFR values less than 0.8. At the interface of the DI core and the PSC jet boundary, station two, there is still a high probability of encountering a mixture with the bulk charge, a probability that at ~2.5-3% is three times higher than that of encountering a stoichiometric mixture. The highest probabilities cover the range of RAFRs from 0.3-0.8, demonstrating that the interface of the DI and PSC jets is still predominantly rich, and if ignitable, capable of producing the robust burn required by partially stratified combustion. As may be expected, the pdfs of RAFR for the DI shear layer (station three, Figure 8-9(c)) and the central portion of the vortex ball of the DI jet (station four, Figure 8-9(d)) are similar, however the similarity is closer than may be expected. Station three shows a slight plateau in probability around the stoichiometric composition.             (a)               (b)              (c)               (d) Figure 8-9 Probability Density Functions for Central Condition of the Test Matrix at (a) Station One, (b) Station Two, (c) Station Three and (d) Station Four. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 143  The only data points that displayed RAFR pdfs that varied from the central condition of the test matrix were those of station one for conditions which corresponded to a) λbulk=1.57 t=-3.53CAD (Figure 8-10(a)) and b) λbulk=1.5 t=0CAD (Figure 8-10(b)). There is a clear reduction in the probability of generating mixtures in the range of RAFR from 0 to 0.4.  There is no ready explanation for this observation since the observed drop cannot be isolated from either the background charge or timing variables, nor is the drop seen in the corresponding bulk charge or timing data-points of the test matrix.                           (a)                (b) Figure 8-10 Probability Density Functions for Capillary Injected PSC Jet with DI at Station One for (a) RAFR=1.573 t=-3.53CAD and (b) RAFR=1.5 and t=0CAD.  The propensity for a mixture to ignite is governed by, amongst other things, its composition. In trying to establish the likelihood of ignition for the PSC jet the same four stations examined for their RAFR pdfs were evaluated for their RAFR minima, maxima and means. Figure 8-11 shows the calculated values for stations one through four ((a)-(d) respectively), for each of the nine test conditions. Station one, the core of the PSC jet, would be expected to yield only pure fuel compositions, yet it is clear from Figure 8-11(a) and the pdf of Figure 8-10(a) that other mixture compositions are also encountered, and that the minima is equal to the background charge strength. The same is true of the other stations, as is the propensity for the fuel maximum concentration to be that of pure fuel, or very close to it. These findings support the work of Schefer et al. [22] who suggest mass engulfment is responsible for gas entrainment. Ahmed & Mastorakos [395] report ignition probability to decrease rapidly downstream of the stoichiometric contour, and with increased dilution. This raises some concerns about igniting the PSC mixture on its shear boundary or at its head vortex, since if its structure is similar to that of the DI jet, these locations have mean mixture RAFRs lean of stoichiometric (Figure 8-11 (c) & (d)). ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 144                      (a)                        (b)                     (c)                       (d) Figure 8-11 Maxima, Minima and Mean RAFR for all Capillary Injected PSC Test Conditions at (a) Station One, (b) Station Two, (c) Station Three and (d) Station Four. The probability density functions, and extrema plots of Figure 8-11 only indentify part of the mixtures’ propensity for ignition. Table 8-2 shows the nine test conditions and the corresponding probability that any given pixel within the 1mm2 is within the flammability limits of methane as specified by [396, 397]. The limits applied from [396, 397] are for a mixture at 1 atm and 250C; however, it is noted by Jessen [397] that, with “the existence of cool flames at elevated temperatures, upper limits often increase drastically and unpredictably with a slight increase in temperature”, which certainly may be the case for the environment within the RCM cylinder, although it is noted by Burgess & Hertzberg [396] that the increase in pressure is unlikely to affect the flammability limits. These observations notwithstanding, it is apparent that at all stations the probability of encountering a mixture within the flammability limits is typically less than 50%, and substantially less at station one for all conditions, although slightly over at station four. Neither the injection timing nor the bulk charge RAFR were found to have a correlation with the probability of the mixture falling within the flammability limits for any of the stations, suggesting that the interaction of the two jets did not significantly alter the entrainment or engulfment behaviour of either ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 145 at these locations. It is desirable to ignite the PSC jet where the composition is rich, since there is a higher probability of successful ignition and a higher ignition kernel growth rate [395], both of which are desirable for successful ignition of the DI jet or homogeneous charge.  Probability Within Flammability Limits λbulk Relative Timing (CAD) Stn. 1 Stn. 2 Stn. 3 Stn. 4 1.5 0 0.32 0.49 0.48 0.50 1.573 3.53 0.30 0.50 0.48 0.54 1.75 5 0.15 0.40 0.57 0.58 1.937 3.53 0.09 0.28 0.52 0.60 1.75 0 0.13 0.36 0.49 0.51 1.937 -3.53 0.11 0.24 0.57 0.55 2.0 -5 0.13 0.37 0.53 0.55 1.573 -3.53 0.22 0.47 0.50 0.51 1.75 0 0.15 0.41 0.55 0.60 Table 8-2 Probability of Mixture Falling Within Flammability Limits for Capillary Injected PSC Jet with DI at Stations One to Four  Despite the limited probability of a flammable mixture at these locations, the mean area over which a rich flammable mixture exists is itself significant, as is evident from Figure 8-12(regions outside the flammability limits of methane are shown in black). The stronger the background charge, the smaller the region within the PSC jet that falls outside the flammability limits (rows viewed from right to left), presumably due to the introduction of the bulk charge which is itself closer to the median flammable concentration. However, Ahmed & Mastorakos [395] note that there is a finite probability of ignition over an area that is larger than the locus of the lean and rich flammability limits, most likely due to the convection of hot gas generated by the spark into a region within the flammability limits. This would suggest an even larger area over which the PSC jet may be ignited. The argument is even extended to show a finite ignition probability in areas where there are no r.m.s concentration fluctuations and the streams may be considered to be of pure fuel or air [398].  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 146  Figure 8-12 Capillary Injected PSC with DI Jet Fuel Concentrations by Volume.    ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 147     Figure 8-13 Probability of Encountering a Mixture Within the Flammability Limits of Methane for Insert Injected PSC  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 148 Figure 8-13 shows the probability of encountering a mixture within the flammability limits of methane based upon the 51 or 153 realisations of each test condition. The most notable observation is that there is a remarkable similarity between the conditions with neither bulk charge concentration nor injection timing having a significant effect upon the probability. There are some small differences in the PSC jet and at the head of the DI jet where it interacts with the PSC jet. The core of both jets exhibit low, but not zero, probability of having a mixture within the flammability limits, while in the head of the PSC jet and within the shear layers of both jets the probability of encountering a flammable mixture is far from guaranteed with probabilities in the range 0.4-0.7. The only regions where there is almost certainty of encountering a flammable mixture are those which would be deemed over mixed, i.e. those at the head of the DI jet.  Fluid motion and chemical kinetics must act together if a flame is to persist or if ignition is to be attained. If the DI jet is to be inflamed and stay so it must not be excessively strained. Turbulent motion ensures that throughout the jets there is an ensemble of strains and strain rates which vary in space and time. Turbulence serves to mix the fuel and oxidizer to attain a combustible mixture, yet the same mixing also serves to dilute the radical pool generated by increased temperatures and transport heat away. The initial chain branching reactions associated with increased temperature are critical to ignition. Methane ignition typically follows the path CH3→CH3O →CH2O→HCO→CO with the kinetically controlled quasi-steady-state species, like HCO, showing a narrow localised concentration profile and the ‘sink’ species, like CH2O, showing a much larger profile [399]. During the ignition delay the chain branching reactions yield a radical pool that grows exponentially with time [119] but for which the slower more spatially distributed species are susceptible to mass transport. Fotache et al. [399] note that “methane/air ignition involves the effects of convective and diffusive losses, as well as chemical heat release”. The interaction of all these variables is complex, as is noted by Grout et al. [400] who observe that at high temperatures (high radical pool concentration), turbulence hinders ignition while at low temperatures (lower radical concentration), turbulence encourages ignition. In the methane/air (H- C-O) system pressure does not change the ignition mechanism [399] since the initial branching reactions do not necessitate a third body. No simple conclusion as to the effect of strain can be made since strain affects inflammation differently at different times. Ahmed et al. [398] note that local straining is detrimental to ignition because the strain and dilation act to reduce temperatures curtailing the radical pool; yet, following the primary stages of the ignition the slower species are affected directly by strain induced mass transport and help propagate the flame. The current LIF set-up cannot ascertain the strain rate, and as such this critical parameter should be sought through computational fluid dynamics (as scalar variance [401]) or through laser velocity techniques.  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 149 “Scalar dissipation describes the rate at which concentration nonuniformities relax towards zero due to molecular diffusion. As such, it provides an important measure of the local rate of mixing at the molecular scale. Knowledge of this mixing rate is particularly important…where chemical reaction rates may be limited by the rate at which reactants are mixed locally” [22].  Scalar dissipation, χ, is calculated according to Eq. 8.1 where Z is the mixture fraction and the binary diffusion coefficient, D, of methane into nitrogen 22mm2/s [22] (a reasonable proxy for the multi-component system of natural gas into air).    Eq. 8.1.  As expected there are high dissipation rates near the PSC origin (Figure 8-14), however, the location of the higher dissipative region later in the jet is not as may be expected. It would be expected, and has been reported  [402], that the dissipation is coincident with the AFR r.m.s fluctuations, and the regions of highest scalar dissipation are not dissimilar to those found in the work of Markides & Mastorakos [178] who looked at the dissipation and mixing of an acetone jet into co-flowing air. It is clear, however, that the dissipation is increased in the vicinity of the PSC, and to such an extent that combustion may not be supported. Numerical work by Rogg et al. [403] on a counter-flowing methane-air flame suggests a critical scalar dissipation rate of 20.6s-1, above which combustion will not occur since convective- diffusive heat removal is increasing while heat generation drops from the reduced reaction rate.  The PSC insert (cf. §4▪4▪1) was designed by Logan [404] to generate better mixing of the PSC injected fuel with the bulk charge and as such reduce the fuel concentration gradients. The current work assesses to what extent this has been achieved, and allows an assessment of how this has been achieved with the aim of informing future insert designs. As was seen previously (cf. Figure 8-4), the insert alters significantly the fuel distribution within the RCM cylinder, and the spatial extent to which the PSC injected fuel interacts with that of the DI jet.    ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 150     Figure 8-14 Scalar Dissipation Rate for Capillary Injected PSC with DI Jet  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 151 8▪5 PSC INSERT INJECTION IGNITION & INFLAMMATION POTENTIAL    Figure 8-15 Insert Inject PSC & DI Jet RAFR Gradients  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 152  Figure 8-16 Insert Injected PSC and DI Jet Normalised RAFR Fluctuations  ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 153 The bulk charge is seen to have two contrasting effects upon the insert injected PSC-DI interaction. The first, and most apparent, is that as with the capillary injected PSC, the more rich the bulk charge the greater the observed reduction in RAFR gradients for the regions, of the DI jet in shear and, in this case, the injected PSC fuel (reduction in observed gradients across rows of Figure 8-15 from right to left). The magnitude of the RAFR gradients in these regions is comparable to those for the capillary injected PSC-DI case. The second observation contrasts that of the shear regions in that the core of the DI jet demonstrates an increase in RAFR gradient with increased bulk charge fuel concentration; this also contrasts the capillary injected PSC case that shows no discernable difference in the RAFR gradients for the DI jet core with varying RAFR (cf. Figure 8-6). These observations are most likely attributable to the fact that the mixing of the DI jet core is unaffected by the PSC capillary injection, and the mixing confined to the jet shear layer, notwithstanding the discussions of mass engulfment. In the case of the insert injected PSC fuel, the injection and subsequent interaction act over a much larger area and, as seen from the mixture distribution images of Figure 8-4, to extend into the core of the DI jet. Furthermore, the lower velocities that the insert generates means that the structure of the insert injection is more susceptible to perturbation and subsequent mixing from the DI jet than the capillary injected jet. Injection timing shows no effect upon the observed gradients (images viewed along colums of Figure 8-15).  In contrast to the capillary injected PSC normalised r.m.s. RAFR flutuations of Figure 8-7, the images of  Figure 8-16 show both timing and background charge to have an effect, while in general the r.m.s. levels are lower with maximum fluctuations about 33% lower. The images for the top row of Figure 8-16 are simlar, as are those for the bottom row. The timings for these rows of images are similar but not identical, with the central images being ±1.47CAD offset from those at the end of the rows (cf. central composite test matrix, Figure 8-1). What is apparent from these images, that is not so apparent from the RAFR images of Figure 8-4, is that the two distinct jets produced by the insert within the interrogation volume (Figure 7-13, p.127) have been mixed into a single larger region. At retarded relative DI injection timings a homogeneous normalised r.m.s. RAFR structure exists with a centre at about x=2-3mm, while for advanced DI injection timings a distinct fluctuating core with a centre at about x=0.5-1mm is seen; there is also seen to be roughly a two-and-a-half fold increase in the flutuation level between the the two timings. This shows that the relative timing of the jets, through momentum exchange and mixing, is able to change the location at which RAFR fluctuations occur- something that was not seen with the capillary injected PSC jet because of the dominance of the PSC momentm over that of the DI jet by the time it reaches the cylinder centreline. As with the capillary injected PSC event, the fluctuations are most prominent in the PSC injected fuel, with the DI jet showing levels about two-thirds lower. This would suggest that when sustaining a flame after spark and its propagation to the DI jet, it is the flutuations and implied strain within the PSC injection not the DI jet that are of concern. The images of Figure 8-16 imply that it is not bulk charge or timing that act to control the fluctuation levels but the combination of the two; the data ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 154 points for retarded DI timings and weaker bulk charge (top right of the test matrix CCD) show lower RAFR fluctuation levels than those for advanced relative DI timing and stronger bulk charge mixtures.  As with the capillary injected PSC-DI interaction, the stochastic nature of the mixture distibution was considered at five locations (Figure 8-17 & Table 2-2), one more than for the capillary inject case. These locations were chosen to assess variation with distance from the firedeck (stations one to three), the point at which the DI jet centreline is interacting with the insert inject fuel (station four) and the location on the downstream side of the DI jet after it has passed through the insert injected fuel (station five).  Figure 8-17 Sample Station Locations for Insert Injected PSC DI Events  Image Position (mm) Station # x y 1 0 2 2 0 4 3 0 6 4 -2.5 6 5 2.5 8 Table 8-3 Insert Injected PSC Sample Station Locations  The probability density function was produced for each set of 13,000 or 39,000 points that corresponded to the 1mm2 area at each location for the 51 or 153 realisations of that test condition, and are presented with a point at the centre of each RAFR interval (set to be 1/48 of the total RAFR range). The results for ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 155 stations one to three are shown in Figure 8-18 for the central row of the test matrix, i.e. a relative injection timing of 0CAD. No data is presented for test conditions at different timings since inspection of the full set of pdfs (shown in Appendix C) displayed little or no dependence up timing. At station one (closest to the firedeck) the three pdfs of Figure 8-18(a) show much the same form, suggesting that the bulk charge has had no chance to mix with the insert injected fuel, and that the RAFR probability distribution is governed by the mixing within the insert prior to the fuel entering the cylinder. At station two (Figure 8-18(b)), 4mm off the firedeck, the bulk charge with the strongest fuel concentration (λbulk=1.5, left-hand image) shows a marked increase in the chance of encountering a mixture with RAFR~0.2, while the case with λbulk=1.75 shows little difference to the case for station one, and at the leanest case only a slight increase in the maximum probability, which occurs at the same RAFR (0.5) as the λbulk=1.75 case. These observations suggest that the mixing mechanisms at work exhibit threshold or stepped behaviour in their propensity to mix the background charge with the injected fuel. Schefer et al. [22] found distinct ‘plateaus’ and ‘ramps’ in the instantaneous concentration distributions for their methane jet, whose effect upon the background charge entrainment may be directly related to the lengthscale of the coherent structures and their associated entrainment. This is supported further by consideration of Figure 8-18(c), the pdfs for station three, which show for the richest background charge a similar distribution to the same condition for station two, although with a reduced peak probability; while the distribution for the weakest bulk charge condition (λbulk=2, Figure 8-18(c) right-hand image) at station three shows little discernable difference from the equivalent condition at station one, 2mm from the firedeck. This ‘matching’ of mixing lengthscales to background charge may also explain the slight increase in the peak probability at RAFR~0.4 for station one and bulk charge RAFR=1.5, although the scatter in the data and the sudden drop in probabilities at RAFR~0.5 make this assertion difficult to fully support. The RAFR distribution at λbulk=1.75 of station three is the only case (of all those for stations one to three) that shows a decrease in peak probability and a plateau in RAFR probability for bulk RAFR values in the range ~0.4-1.2. Of final note is that the probability distributions for stations two to three (Figure 8-18(a) & (b)) show significantly less local variability in RAFR. The differences in probability from one RAFR value to another change gradually, while those at the low and high bulk RAFR values for station one show more scatter. This may be a result of the mixing process which takes place within the insert having started, but not completed, prior to mixing with the bulk charge. In the cases of the stations further from the firedeck the mixing process is more advanced and less time/location dependant. It should, however, be noted that the case for λbulk=1.75  at station one (Figure 8-18(a), centre) does not show the same scatter as the two other cases (Figure 8-18(a), left and right). There is no immediately clear reason for this difference, although it is worth noting that the work of Gorby used a bulk charge RAFR of 1.79 (very close to the 1.75 condition) and a spark plug which would produce a PSC jet more akin to that of the insert than the capillary and found more reliable ignition at this condition. The lack of scatter in the RAFR pdf may further point to the bulk charge smoothing RAFR fluctuations rather than smoothing concentration ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 156 gradients, particularly if the lengthscales of the mixing process match that required by the bulk charge to produce a ‘desirable’ ignition condition (with the details of what constitutes a ‘desirable’ condition still in need of exact definition).  Figure 8-18 RAFR Probability Density Functions for t=0CAD at λbulk=1.5 (left), λbulk=1.75 (centre) and λbulk=2 (right) for (a) Station One, (b) Station Two and (c) Station Three.  Stations four and five exhibit pdfs that are markedly different to those of stations one to three and are markedly different to each other despite what would appear from the RAFR images of Figure 8-18 to be similar regions. Station four (Figure 8-19(a)) shows a reduction in bulk RAFR to move the peak probability in local RAFR from ~0.7 to ~1, for a change in bulk RAFR of 1.5 to 2. Furthermore, the magnitude of the peak probability is seen to increase from ~3% to ~5.75%, increasing the kurtosis (‘peakyness’) of the distribution correspondingly. The distribution for station four with λbulk=2 demonstrates a significant probability that the bulk charge concentration will still be present, a probability that, at 3.75%, is comparable to encountering RAFR values of ~0.5 or ~0.9. This suggests that the ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 157 engulfment mechanisms proposed by Schefer et al. [22] are still present and that their extent, and the mixing they cause, is dependent upon bulk charge fuel concentration. Station five (Figure 8-19(b)), at the front of the DI jet after it has passed through the insert injected fuel, shows less marked variations with bulk RAFR changes. The two leanest bulk charges (λbulk=1.75 & 2) show similar forms to their pdfs, although at λbulk=1.75 the distribution rises rapidly to a peak probability of ~1.5% at a bulk RAFR of ~0.5, dropping by only ~0.5% over the range λbulk=0.5-1.75. With a bulk RAFR of two, the peak probability of ~2.5% extents over the range λbulk=0.4-0.8, before dropping steadily to a probability of ~1% at bulk RAFR, where no sudden increase in probability is observed as for station four. The most rich bulk RAFR (1.5) is observed to cause a peak probability of ~2.5% in the local RAFR at 0.3, but which is not sustained and falls to a plateau at ~1% over the range of ~0.5-1.5.   Figure 8-19 RAFR Probability Density Functions for t=0CAD at λbulk=1.5 (left), λbulk=1.75 (centre) and λbulk=2 (right) for (a) Station Four and (b) Station Five.  Figure 8-20 RAFR Probability Density Functions for t=-5CAD (centre) and t=-3.53CAD (left & right) at λbulk=1.57 (left), λbulk=1.75 (centre) and λbulk=1.93 (right) for Station Five.  (a) (b) ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 158 For station five, at the two most advanced injection timings (t=-3.53 & -5CAD) (Figure 8-20), the RAFR probability distributions are a lot less variable than those at more retarded timings, showing a gradual increase in local RAFR from 0.1 to the bulk RAFR. The bulk RAFR is seen to be the maximum RAFR for λbulk=1.5, while for the leaner bulk RAFR cases the bulk charge serves to increase the probability of richer local mixtures, although lean of stoichiometric. The trends in the pdfs of Figure 8-20(b) demonstrate once more the need to match bulk RAFR with timing and the level of mixing that any given timing engenders.  Figure 8-21 shows some similar features to that of the capillary injected insert with respect to the minimum, mean and maximum RAFR that are encountered at stations one to three (Figure 8-21(a)-(c) respectively). In all cases the maximum RAFR is that of the bulk charge, while the mixing that takes place within the PSC insert raises the mean RAFR slightly to ~0.5-1. The exception is station one where, in the cases for λbulk=1.5 & 2, mixing has failed to fully eliminate packets of pure fuel. The need to match the bulk charge RAFR to the level of mixing at the desired spark location is once again clear, with station one showing, on average, rich mixtures for bulk charges of λbulk=1.5 & 1.57, while for λbulk=1.75, 1.93 & 2, the mean condition is close to stoichiometric. The opposite is true of station two where, from bulk RAFR=1.5-1.75, the mean RAFR is approximately stoichiometric while, for bulk RAFRs higher than 1.75, the mean condition is rich. Investigation of the station’s RAFR dependence on injection timing showed no correlation, although it is apparent from Figure 8-18(c), which corresponds to station three, that there is still a large range, spanning 0.8 to 1.25 at λbulk=1.75, in the expected value of RAFR for the 1mm2 of the station. This may be attributable to the fluctuations still present in the insert injected fuel RAFR being increased by those associated with the DI jet fluctuation maximum, which occurs slightly off-axis (cf.§2▪2▪1).   (a)         (b)             (c) Figure 8-21 Maxima, Minima and Mean RAFR for all Insert Injected PSC Test Conditions at (a) Station One, (b) Station Two and (c) Station Three  Stations four and five (Figure 8-22 (a) & (b) respectively) show maxima and minima in line with those of station one to three. Station four demonstrates mean RAFR variations similar to station one, although no direct link between the two is thought to exist. The range of mean RAFRs for station four ranges from ~0.8 ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 159 to ~1.2, with the leanest cases all at bulk RAFR of 1.75. At station five the mean RAFR is consistently on the lean side of, or close to, stoichiometric, covering the range ~1 to ~1.5. The data does not support an effect from the bulk RAFR, with the station RAFR appearing almost independent of bulk RAFR, and more likely to be driven by the mixing which takes place upstream as the fuel from the DI jet interacts with that of the PSC insert.                     (a)                       (b) Figure 8-22 Maxima, Minima and Mean RAFR for all Insert Injected PSC Test Conditions at (a) Station Four and (b) Station Five  Figure 8-23 shows the mean RAFR against the relative injection timing for stations one and five (no other stations showed a dependency upon injection timing, nor did any on bulk charge RAFR). An increase from a mean RAFR of ~0.8 to ~1.3 with a 10CAD advance in relative DI injection timing (regression coefficient 0.575) was observed at station five, while the same timing change caused a decrease in mean RAFR from ~1.55 to ~1 for station five (R2=0.831). It can be seen from this that the mixing generated by the interaction of the PSC and DI jets has distinctly different effects upon different locations within the cylinder and, as such, mixing must be considered in conjunction with the specific location of the spark. The timing must be assessed in combination with an assessment of the desired local RAFR, and the extent to which a rich mixture is desirable for its increased combustion strength.  For the insert injected PSC jet, the probability that any given pixel within the test station area (for all realisations) will fall within the flammability limits of methane for the nine test conditions is shown in Table 8-4. Stations one to three show similar trends with the highest probabilities of a flammable mixture being found at about λbulk=1.75, with a 6-10% increase in the probability from station one to three. The data on bulk charge effect did not support more firm conclusions beyond this, with the most apparent second order polynomial returning regression coefficients less than 0.5 for all three stations. Injection timing offered no demonstrable effects upon forming a mixture that falls within the flammability limits. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 160  Figure 8-23 Effect of Relative injection Timing on Mean RAFR for Station One (left) and Station Five (right)  The increase in the probability of the mixture falling within the flammability limits of methane was calculated for stations one and five, since they are at locations comparable to those of the capillary injected PSC-DI event. It is clear from Table 8-5 that the increase is far from uniform for the different bulk charge RAFRs and injection timings, and showed no correlation to either timing or bulk charge. The two locations do show different general responses however, with station one (immediately after the point of PSC injection) seen to exhibit two-and-a-half to six-fold increases in the probability of encountering a flammable mixture for the cases, in which λbulk=1.75-2, while at low background RAFR the increases are small (13%-95%). At station five, the point after the DI jet has passed through the PSC jet, the insert is seen to offer only very small increases in the probability of encountering a flammable mixture (3-8%) for bulk RAFR of 1.75 to 1.93, while in all other cases there is a reduction in the probability of 9-37%.  Probability Within Flammability Limits λbulk Relative Timing (CAD) Stn. 1 Stn. 2 Stn. 3 Stn. 4 Stn. 5 1.5 0 0.36 0.27 0.38 0.61 0.31 1.573 3.53 0.43 0.47 0.50 0.67 0.49 1.75 5 0.54 0.61 0.64 0.69 0.50 1.937 3.53 0.63 0.61 0.58 0.56 0.62 1.75 0 0.57 0.52 0.63 0.65 0.55 1.937 -3.53 0.44 0.39 0.53 0.59 0.57 2.0 -5 0.34 0.31 0.57 0.73 0.40 1.573 -3.53 0.43 0.37 0.38 0.68 0.32 1.75 0 0.51 0.48 0.52 0.71 0.40 Table 8-4 Probability of Mixture Falling Within Flammability Limits for Insert Injected PSC Jet with DI at Stations One to Five   ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 161 Increase in Probability Station Mixture is Within Flammability Limits Stn. 1 Stn. 5 λbulk Relative Timing (CAD) --- % Increase --- % Increase 1.5 0 0.04 13% -0.19 -38% 1.573 3.53 0.13 43% -0.05 -9% 1.75 5 0.39 260% -0.08 -14% 1.937 3.53 0.54 600% 0.02 3% 1.75 0 0.44 338% 0.04 8% 1.937 -3.53 0.33 300% 0.02 4% 2.0 -5 0.21 162% -0.15 -27% 1.573 -3.53 0.21 95% -0.19 -37% 1.75 0 0.36 240% -0.2 -33% Table 8-5 Increase in Probability of Mixture Falling Within Flammability Limits for Insert Injected PSC Jet Over Capillary Injected Jet at Stations One and Five  It can be seen from Figure 8-24 that the injected mean fuel concentration lies almost wholly within the flammability limits of methane, and is in the upper half of the range. The regions of lower volume fraction are confined to the weak bulk charge; even the shear layer of the DI jet shows fuel volume fractions of ~ 10%, while the spatial extent of those regions is constant at ~1mm regardless of injection timing or bulk charge RAFR. In all cases the PSC injected fuel appears to have similar volume fraction distributions, namely an inner core with fuel concentrations in the range 10-13%, reducing at the edges to ~8%. Of most note is the effect of the bulk charge RAFR upon the core of the DI jet. In the cases with RAFR from 1.5 to 1.57 (left-hand column of Figure 8-24) the DI core is fully within the flammability limits while, as the bulk charge is leaned, the extent of the core that falls outside the flammability limits increases (more black in right-hand column than centre). This seems at first counter-intuitive, the mixing in of a more dilute mixture should dilute the DI core more than a richer bulk charge, as was seen with the capillary injected PSC & DI combination. In the capillary injected PSC case, it was noted that the PSC injection maintained its structure and had little effect upon the DI jet upstream of the injection point. In the case of the insert injected PSC jets the mixing is over a larger area and affects the mixing and entrainment that the DI jet undergoes. It is, therefore, necessary to consider the mechanisms driving the mixing. The larger concentration gradient between the jet core and the bulk charge at high bulk RAFRs would drive greater diffusion; yet, since air has both a higher viscosity and molecular weight than methane, it is reasonable to expect that at higher RAFRs there is less viscous mixing taking place. The data from the current study are not able to support a comprehensive assertion as to whether this is the case, and assessment of the local Prandtl number would need to be made, which could be done through either ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 162 computational fluid dynamics or a combination of laser based experiments to assess local fluid motion and temperature (which would then need to be related to the diffusivity of methane into air). Consideration of the flow field would also then allow an assessment of the local strain rate, an important factor in both ignition and flame propagation, and its relation to the local scalar diffusion rate.  Figure 8-25 shows the probability of encountering a mixture within the flammability limits of methane, and it shows some clear differences to that of the capillary injected case (cf. Figure 8-13). The less strong the bulk charge to lower the probability of encountering a flammable mixture in the core of the DI jet, the probabilistic data confirming the observations of the mean case (Figure 8-24). The mixed region of the PSC and DI jets shows probabilities which relate closely to the mean mixture concentration. The higher the concentration the higher the probability of having a flammable mixture, although there is no clear direct correlation. In all cases the mixed region has probabilities of being with the flammability limits in the range 0.5-0.8, although where there is still a richer core present (the three cases of λbulk=1.5, and t=-3.53 with λbulk=1.57 & λbulk=1.93) within the mixed region this probability drops to 0.4-0.5. In general the insert has been seen to raise the chance of encountering a flammable mixture from  ~0.5 to ~0.8, though more significantly the are of which a flammable mixture can be expected is an order of magnitude larger. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 163  Figure 8-24 Insert Injected PSC with DI Jet Fuel Concentrations by Volume. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 164     Figure 8-25 Probability of Encountering a Mixture Within the Flammability Limits of Methane for Insert Injected PSC   ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 165 As was seen from the scalar dissipation data for the capillary injected PSC, ignition was unlikely since the dissipation rate was above that which can support combustion. However, the insert is shown in Figure 8-26 to lower the scalar dissipation rate to values well below the critical extinction value of 20.6s-1 [403], and values which are an order of magnitude lower than the capillary injected PSC jet. The difference is large enough that the dissipation rate may be too low to support a robust combustion, and is in need of more investigation. The dissipation, as may be expected, has a spatial distribution almost identical to that of the r.m.s. fluctuations [402]; however, it must be cautioned that the out-of-plane component of the scalar dissipation rate cannot be assessed. This was of little concern in the capillary injected jet since, if desired, it could be assessed to be similar to that in x-direction, yet for the insert injected fuel there is no way to ascribe a value to the out of plane component. To circumvent this problem, a series of images across the section of the injected fuel would need to be made, and on two planes. ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 166     Figure 8-26 Scalar Dissipation Rate for Capillary Injected PSC with DI Jet ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 167 8▪6 CONCLUSIONS Chapter 8 has presented the findings of repeating the test matrix presented in Chapter 7 with a smaller image area that contains more detail of the PSC-DI interaction, and with higher fidelity LIF signals. Comparison has been made between a PSC injection delivered from a capillary tube and one from a PSC sparkplug ‘insert’.  The interaction of the DI jet with that of the capillary injected PSC jet is seen to maintain the structure of the PSC injection and perturb the DI jet only locally to the PSC jet. The mixing is still confined to shear layers of the DI jet and is unaffected by the bulk charge RAFR, yet the bulk RAFR is seen to alter the mixture distribution in the head structure of the DI jet, increasing the area covered by an approximately stoichiometric composition. In contrast to the capillary injected PSC jet, that of the insert jet was readily altered by the DI jet, since the momentum of the insert injection was lower than that of the capillary injection, resulting in the individual insert jets being amalgamated into a single almost homogeneous region. The head vortex structure of the DI jet was almost fully supplanted with the amalgamated distribution. The quasi-homogeneous mixture had a composition close to stoichiometric and may burn vigorously enough to inflame a lean bulk mixture or the DI jet, although this depends on a wide range of local and global in-cylinder conditions, including strain. The insert showed, on average, no discernable differences in structure as a result of the approaching DI jet, yet for some individual realisations the capillary injected PSC jet showed clear deflections.   The potential for the capillary injected PSC to be ignited was considered with respect to the fuel concentration gradients. The shear layer in the DI jet was an order of magnitude thicker than that of the PSC jet, although it was proposed that the effects of the background charge was not solely to alter the concentration gradients but to also reduce the fluctuations in RAFR, increasing the likelihood of encountering a flammable mixture. This is of particular importance, since the work of Schefer et al. [22] has been drawn on to provide insight into the entrainment mechanisms at work. The entrainment of the bulk charge was not likely to be uniform or steady; mass ‘engulfment’ being more likely with the coherent structures giving rise to large local variations in mixture AFR that may vary significantly from the experimental average condition. The probability density functions for a range of stations within the mixture were presented, all of which displayed the full range of RAFR values from pure fuel to bulk mixture concentration, even for locations where the experimental average suggested these concentrations were not possible. In many cases the overall probability of encountering a flammable mixture throughout the area of the station was less than 50%. There was found to be a marked similarity in the RAFR pdfs for the locations at which the DI centreline interacts with the PSC jet and the DI head vortex after it had passed through the PSC jet. Both relative injection timing and bulk RAFR were found to have little effect upon the ▪ Chapter 8 ▪ The Interaction of Partially Stratified & Direct Injection Fuel Jets ▪ 168 mixture distribution. The discuss