An experimental and model ing study of the ignition of dimethyl carbonate in shock tubes and rapid compression machine

Please note that changes made in the online proofing system will be added to the article before publication but are not reflected in this PDF.

We also ask that this file not be used for submitting corrections.

ContentslistsavailableatScienceDirect

Combustion and Flame

journalhomepage:www.elsevier.com/locate/combustflame

An experimental and model ing study of the ignition of dimethyl carbonate in shock tubes and rapid compression machine

Katiuska Alexandrino

^a,b,∗

, María U. Alzueta

^b

, Henry J. Curran

^a

Q1

a Combustion Chemistry Centre, School of Chemistry, NUI Galway, Ireland

b Aragón Institute of Engineering Research (I3A), Department of Chemical and Environmental Engineering, University of Zaragoza, 50018 Zaragoza, Spain

a rt i c l e i n f o

Article history:

Received 14 July 2017 Revised 22 August 2017 Accepted 2 October 2017 Available online xxx Keywords:

Shock tube

Rapid compression machine Dimethyl carbonate Oxidation Kinetic model

a b s t r a c t

Ignitiondelay timesof dimethyl carbonate DMC weremeasured usinglow- and high-pressure shock tubesand inarapidcompressionmachine(RCM). Inthisway, theeffectoffuelconcentration(0.75%

and1.75%), pressure(2.0, 20,and 40atm)andequivalenceratio(0.5, 1.0,2.0)onignitiondelaytimes wasstudiedexperimentallyandbymodeling.Experimentscoverthetemperaturerangeof(795–1585K).

Severalmodels from literature wereused to performsimulations, thus theirperformances to predict thepresentexperimentaldatawasexamined.Furthermore,theeffectofthethermodynamicdataofthe CH3O(C=O)˙OradicalspeciesandthefuelconsumptionreactionCH3O(C=O)OCH3  CH3O(C=O)˙O+CH3, onthesimulationsoftheignitiondelaytimesofDMCwasanalyzedusingthedifferentmodels.Reaction pathand sensitivityanalyseswerecarriedoutwiththefinalmodel topresentanin-depthanalysisof theoxidationofDMCunderthedifferentconditionsstudied.Thefinalmodel usedAramcoMech2.0as thebasemechanism andincluded aDMC sub-mechanismavailableinliterature towhichthe reaction CH₃O(C=O)OCH₃  CH3O(C=O)˙O+CH₃ wasmodified.Goodagreementisobservedbetweencalculated andexperimentaldata.Themodelwasalsovalidatedusingavailableexperimentaldatafromflowreac- torsand opposedflowdiffusionand laminarpremixed flamesstudies showinganoverallgood perfor- mance.

© 2017TheCombustionInstitute.PublishedbyElsevierInc.Allrightsreserved.

1. Introduction 1

Dimethyl carbonate (CH₃O(C=O)OCH₃, DMC), a non-toxic and Q2

2

non-corrosive carbonate esterwith noC–C bonds and containing 3

53%oxygenbyweight,hasbeenidentifiedasasuitablefuelcom- 4

pound tobeaddedtodieselfueltoreducePMemissionswithout 5

affecting NO_X emissions[e.g.,1].Eventhough it is100%miscible 6

with diesel fuel,it must be used asa blendedfuel in diesel en- 7

gines due to its low cetane number (35–36), low calorific value 8

(15.78MJ/kg),andhighlatentheatofevaporation(369kJ/kg)[2,3]. 9

To contribute tothe developmentof detailedchemical kinetic 10

modelstodescribethecombustioncharacteristicsofDMC,athor- 11

oughunderstandingofits combustionchemistryisneeded.These 12

chemical kineticmodels canbe usedinconjunctionwithcompu- 13

tational fluid dynamics (CFD)codes, with the necessary simplifi- 14

cations, to simulate the physical and chemical processes in en- 15

gines,leadingtooptimalengineefficiencywithminimalemissions.

16

∗ Corresponding author at: Aragón Institute of Engineering Research (I3A), De- partment of Chemical and Environmental Engineering, University of Zaragoza, C/ Mariano Esquillor, s/n , 50018 Zaragoza, Spain.

E-mail address: [email protected] (K. Alexandrino).

Tothisend, studies addressing thethermal decomposition[4–6], 17 photolysis[7]andoxidationofDMChavebeenreportedinthelit- 18

erature. 19

SinhaandThomson[8]measuredspeciesconcentrationsacross 20 DMC/air and propane/DMC/air opposed flow diffusion flames. 21 Formaldehydewasfoundtobean importantintermediatespecies 22 intheDMCflame,andthepresenceofoxygenonthecentralcar- 23 bon in DMCfavors breakageof the O–CO bond which results in 24 very low levels of formation of methane, ethane, ethylene, and 25 acetylene. Glaude et al. [9] developed the first chemical kinetic 26 sub-mechanismforDMCconversion,whichwasincorporatedinto 27 apreviouslydevelopedchemicalkineticmechanismfordimethoxy 28 methane(DMM)anddimethylether (DME)[10,11]. Thepredicted 29 composition profiles using this model were in reasonable agree- 30 mentwiththemeasuredspeciesprofilesfromSinhaandThomson 31

[8]. 32

Chen et al. [12] investigated the oxidation of 33 n-heptane/DMC/O₂/Ar mixtures in a laminar pre-mixed low- 34 pressure (30Torr) flame, at an equivalence ratio of 1.16, using 35 synchrotronphotoionizationandmolecular-beam massspectrom- 36 etry(PI-MBMS)techniques.Measuredandsimulatedmolefraction 37 profiles of major and intermediate species were compared. The 38 https://doi.org/10.1016/j.combustflame.2017.10.001

0010-2180/© 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Pleasecitethisarticleas:K.Alexandrinoetal.,Anexperimentalandmodelingstudyoftheignitionofdimethylcarbonateinshocktubes

calculations were performed using a model which includes the 39

DMM and DMC sub-mechanisms from the model developed by 40

Glaude et al. [9]. The predicted concentrations of flame species 41

agree reasonably well with the measured results. The authors 42

observed an early production of CO₂ in both the measured and 43

modeling results which is suggested to occur mainly dueto the 44

decompositionofmethoxyformyl(CH₃O˙C=O)radicals.

45

Peukertetal.[13,14]studiedthehightemperaturethermalde- 46

compositionofDMC,anditsinteractionwith ˙HandÖ atoms,us- 47

ingtheshocktubetechniqueinconjunctionwithmasterequation 48

analysis.Bardinetal.[15]measuredlaminarburningvelocitiesof 49

DMC/air flames at initial gas mixture temperatures of 298, 318, 50

338,and358K.Theseresultswere simulatedusingthemodelde- 51

velopedby Glaudeetal.[9],andit wasfound thatthemodelsig- 52

nificantlyover-predictedthemeasuredlaminarburningvelocities.

53

More recently, Hu et al. [16] measured ignition delay times 54

of DMC oxidation in a shock tube at high temperatures (1100– 55

1600K), at different pressures (1.2–10 atm), fuel concentrations 56

(0.5–2.0%)andequivalenceratios(

ϕ

=0.5–2.0).Achemicalkinetic 57

model,basedonthemodificationoftheDMCsub-mechanismfrom 58

Glaude etal. [9] and the AramcoMech 1.3 mechanism [17], was 59

proposed to describe theignition delaytimes of DMC.The mea- 60

suredignitiondelaytimesfromthiswork,aswellastheDMC/air 61

opposeddiffusionflamedatareportedbySinhaandThomson[8], 62

werecomparedwithmodelcalculationsshowinggoodagreement.

63

Sunetal.[18]investigatedbothpyrolysisofDMCinaflowre- 64

actoratdifferentpressures (40,200 and1040mbar)andits oxi- 65

dationin laminarpremixedlow-pressure DMC/O₂/Arflames with 66

equivalenceratiosof1.0and1.5,at25and30Torr,respectively.A 67

detailedkinetic modelfor DMC pyrolysis and combustion,based 68

on a new sub-mechanism for DMC conversion and the Aram- 69

coMech1.3mechanism [17], wasproposed. Thismodel wasvali- 70

datedusingtheexperimentaldataobtainedbySunetal.[18]and 71

withtheopposedflowflame[8],burningvelocities[15]andshock 72

tube[16] experimental data fromthe literature. A more detailed 73

descriptionofthesethreeDMCsub-mechanisms[9,16,18]isfurther 74

showninSection3. 75

Most recently, Alzueta et al. [19] carried out an atmospheric 76

flow reactor study of DMC oxidation in the absence and pres- 77

enceofNOinthetemperaturerange700–400Kat

ϕ

=0.028,1.00, 78

1.43, and 3.33. It was found that, in the DMC–NO interaction, 79

thefuel-rich conditionscontributeslightlytothenetreduction of 80

NOx.Adetailedkineticmodel,basedontheDMCsub-mechanism 81

fromGlaude etal. [9]and a core mechanismdescribed andup- 82

dated by the authors [20–24], was proposed. With this model, 83

these authors evaluated the impact of the thermodynamic data 84

on the modeling results, finding that the enthalpy of formation 85

oftheso-calledDMCradical CH₃O(C=O)˙O significantlyinfluences 86

theDMCconversionresults.The influenceofthethermodynamics 87

ofthe CH₃O(C=O)˙Oradicals speciesonourcurrentignitiondelay 88

timemeasurementswillbefurtherdiscussedinSection4.Alzueta 89

etal.[19]alsoperformedoneDMCpyrolysisexperimentinanat- 90

mosphericflowreactorinordertodetermineitscapacitytoform 91

soot,whichwasfoundtobeverylow.Followingthiswork,Alexan- 92

drinoetal.[25]studiedthesootingpropensityofDMCthroughpy- 93

rolysisexperimentsinan atmosphericflowreactorinthetemper- 94

aturerange 1075–1475K andinlet DMC concentrationsof 33,333 95

and50,000ppm.It wasconfirmedthat DMChas avery low ten- 96

dencyto formsoot, even when compared with ethanol, because 97

theformationofCOandCO₂isfavored,andthusfewcarbonatoms 98

are available for soot formation. The formation of CO₂ is highly 99

favoredby the decompositionofthe CH₃O˙C=OandCH₃O(C=O)˙O 100

radicals.Sootreactivityandcharacterizationbyinstrumentaltech- 101

niqueswasalsoconsidered, showingthatthe highertemperature 102

andtheinletDMCconcentration ofsootformation,thelowerthe 103

reactivityofthesoot.

104

Table 1

Composition of DMC mixtures studied in the low-pressure shock tube.

Mix. P 5a(atm) ϕ DMC (mole%) O 2 (mole%) Ar (mole%)

1A 2 0.5 0.75 4.5 94.75

2A 2 1 0.75 2.25 97

3A 2 2 0.75 1.12 98.12

4A 2 0.5 1.75 10.5 87.75

5A 2 1 1.75 5.25 93

6A 2 2 1.75 2.62 95.62

a P 5 is the pressure behind the reflected shock wave.

Toourknowledge,todatetheworkofHuetal.[16]isunique 105 instudyingignitiondelaytimesforDMCoxidation.Althoughthat 106 workcovereda widerangeofequivalenceratios,DMCconcentra- 107 tions and pressures, more experimental data for the ignition of 108 DMC are needed to develop and validate chemical kinetic mod- 109 elstoaccurately describeDMCcombustion.Keeping thisinmind, 110 the aims of this work are: 1) to study ignition delay times of 111 DMCundernewexperimentalconditions,andtherebyextendthe 112 available experimental data of DMCoxidation. In particular, low- 113 andhigh-pressure shocktubes were used and, forthe first time, 114 a rapid compression machine was used in order to extend the 115 high-pressureshocktube dataof DMCto lower temperatures; 2) 116 tocomparetheperformanceofthedifferentmodelsforDMCcon- 117 versionavailableinliteraturetopredictthemeasuredignitionde- 118 lay times ofthis work. This also includes the analysisof the ef- 119 fect ofchanging the thermodynamic ofthe CH₃O(C=O)˙O radicals 120 species and reaction kinetics on the modeling calculations. The 121 goal isto find a modelthat best predicts our experimental igni- 122 tion delay times and also various other experimental targets in- 123 cludingflowreactorsandopposedflowdiffusionandlaminarpre- 124 mixedflames;and3)toperformthechemicalinterpretationofthe 125 effectoftheDMCconcentration,pressureandequivalenceratioon 126 measuredignitiondelaytimesthroughrateofproductionandsen- 127

sitivityanalyses. 128

2. Experimental 129

Shocktubeandrapidcompressionmachineignitiondelaytime 130 datawereobtainedusingthefacilitiesattheNationalUniversityof 131 IrelandGalway(NUIG). Thefull listofmixturesstudied andtheir 132 compositionsareprovidedinTables1and2.Experimentswithsto- 133 ichiometricandfuel-richmixturescompressedat40atmcouldnot 134 be performedin therapid compressionmachine dueto thehigh 135 rateofheat releaseduringignition, whichdamaged the pressure 136 transducerusedtomonitorthepressure. 137 DMCliquid(99%pure,Sigma-Aldrich)wasusedwithoutfurther 138 purification.O₂,ArandN₂cylindersweresuppliedbyBOCathigh 139 purity(99.5%).Mixtureswerepreparedinanevacuatedandheated 140 (348K)stainlesssteelmixingtankateachfacility,usingthepartial 141 pressure method. The fuel wasinjected via an injectionport on 142 thetank usingagas-tightsyringe,followedbythe additionofO_{2 143} and finally the diluent (Ar orN₂). Forall conditions, the partial 144 pressureofthefuelwasmaintainedatavaluelessthanone-third 145 that of its saturatedvapour pressure at the tank temperature to 146 avoidfuelcondensation.Eachmixturewasallowedtohomogenize 147 overnightbeforeuse.Theuncertaintyinmixtureconcentrationsis 148 estimatedtobe± 2%.Theexperimentalprocedureandfacilitiesto 149 determinetheignitiondelaytimedataareexplainedbrieflybelow. 150

2.1. Low-pressureshocktube 151

Tomeasure theignition delaytime ofDMC/O2/Armixtures at 152 high temperatures (T=1220–1585K) and at low pressure (P=2 153 atm) (mixtures in Table 1), a low-pressure shock tube (LPST), 154

Fig. 1. Endwall pressure and CH ^∗ emission traces with the definition of ignition delay time in the low-pressure shock tube. Example for 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), T = 1303 K.

described previously by Smith et al. [26], was used. Briefly, the 155

stainless steel tube consistsof a 63cm long and52cm diameter 156

barrel-shaped driver section, which is coupled to a 6.22m long 157

driven section (internal diameter of 10.24cm). The two sections 158

are separated using a polycarbonate diaphragm. Helium (99.99%, 159

BOC Ireland) was used as the driver gas. Five PCB 113B24 pres- 160

suretransducers,locatedinthedrivensection,wereusedtodeter- 161

mine the shockvelocity at the endwall, takingaccount of shock 162

wave attenuation by linearly extrapolating the five velocities to 163

the endwall. The endwall pressure was detected using a Kistler 164

(603B) pressure transducer. Light emission, which was emitted 165

from excited CH^∗ radicals, was detected using a photo-detector 166

(Thorlabs Inc. PDA55-EC)and a narrow band-pass filter centered 167

at430nm, through a fusedsilica windowembedded inthe end- 168

wall. All pressure and CH^∗ emission signals were recorded using 169

two Handyscope HS4 digital oscilloscopes. The chemical equilib- 170

rium program (GasEq) [27] was used to calculate the reflected 171

shocktemperature (T₅) andpressure (P₅) ofeach shockknowing 172

the test gascomposition, shockwave velocity, initial pressure of 173

testgasandinitialtemperatureofthegas.Theuncertaintiesofthe 174

reflectedshocktemperatureisestimatedtobe± 1K.

175

The ignition delay time was defined as the time interval be- 176

tweentheriseinpressureduetothearrivaloftheshockwave at 177

theendwallandthatduetofuelignition.Fuelignitionisvisibleby 178

followingtheincreaseinlightemission(inthiscaseinformofCH^∗ 179

emission)duetotheignition event.Inthisway,onsetofignition 180

wasdefinedbyextrapolatingthemaximumslopeofCH^∗emission 181

tothebaseline,asindicatedinFig.1.Theuncertaintyinthemea- 182

suredignitiondelaytimeswasdeterminedtobe± 15%,duetothe 183

uncertainties in the conditions behind the reflected shock wave.

184

Fig. 2. Pressure trace with the definition of the ignition delay time in the high- pressure shock tube. Example for ϕ= 1, P = 40 atm (mixture 5B in Table 2 ), T = 1199 K.

Uncertaintiesinthemolefractionsofreactantsareminimal(<5%) 185 ashighaccuracydigitalpressuregaugeswereused. 186

2.2.High-pressureshocktube 187

Theignition delaytimesof DMC/airmixtures athightemper- 188 atures (T=950–1400K) and high pressures (P=20 and 40 atm) 189 (mixtures 1B–6B in Table 2) were measured in a heated high- 190 pressureshocktube(HPST)withdriveranddrivensectionlengths 191 of3.0and5.7m,respectively,andaninternaldiameterof63.5mm, 192 describedindetailpreviously [28].A3cmdouble-diaphragmsec- 193 tionseparatedthe driverandthedriven sections.Pre-scoredalu- 194

miniumplateswereusedasdiaphragms. 195

Thedriversectionwaspressurizedwithpureheliumorwitha 196 helium-nitrogenmixture,thelaterto achieve atailoredcondition 197 inorderto obtainlonger test time[29,30].Approximately halfof 198 thetotaldriverpressurewasfilledintothemiddle-section,acting 199 asabuffer betweenthe muchlower initial testgas pressureand 200

themuchhigherdrivergaspressure. 201

Sixpressuretransducers(PCB113B24),mountedflushtothein- 202 teriorwall of the heateddriven section, were used to determine 203 the shock velocity at the endwall. In the same way, as that in 204 thelow-pressureshocktube,thisvalue wasusedto calculatethe 205 temperatureandpressure behindthe reflected shock wave using 206 GasEq [27]. The pressure at the driven section endwall, used to 207 measure the ignition delay time, was monitored using a Kistler 208 603Bpressuretransducer.Pressuretraces(exampleinFig.2)were 209 obtainedusingtwo HandyscopeHS4 digitaloscilloscopes.The ig- 210 nition delay time wasdefined as the time interval between the 211 pressurerisedueto thearrival oftheshockwave attheendwall 212 andthemaximumrateofpressureriseduetotheignitionevent. 213

Table 2

Composition of DMC mixtures studied in the high-pressure shock tube (HPST) and rapid compression machine (RCM).

HPST RCM

P 5 or P ^Ca(atm) ϕ Mix. DMC (mole%) O 2 (mole%) N 2 (mole%) Mix. DMC (mole%) O 2 (mole%) Ar (mole%)

20 0.5 1B 3.38 20.30 76.32 1C 3.38 20.30 76.32

1 2B 6.54 19.63 73.82 2C 6.54 19.63 73.82

2 3B 12.28 18.43 69.29 3C 12.28 18.43 69.29

40 0.5 4B 3.38 20.30 76.32 4C 3.38 20.30 76.32

1 5B 6.54 19.63 73.82 – – –

2 6B 12.28 18.43 69.29 – – –

a P 5 is the pressure behind the reflected shock wave (in the HPST) and P c is the compressed gas pressure (in the RCM).

Fig. 3. Pressure trace with the definition of the ignition delay time in the rapid compression machine. Example for ϕ= 1, P = 20 atm (mixture 2C in Table 2 ), T = 795 K.

Experimentaluncertaintyinignitiondelaytimeswasestimatedas 214

± 15%.

215

2.3.Rapidcompressionmachine 216

Ignition delay times atlow temperatures (T=795–975K) and 217

highpressures (P=20 and 40 atm) (mixtures 1C–4C in Table 2) 218

weremeasured in arapid compressionmachine (RCM) described 219

previouslybyDarcyetal.[31].Thismachine hasatwinopposed- 220

piston configuration, resulting in a fast compression time of ap- 221

proximately 16ms. Creviced piston heads were used to suppress 222

the in-cylinder roll-up vortices within the combustion chamber 223

andthus to improve the post-compression temperaturedistribu- 224

tioninthechamber.Compressedgastemperatureswerevariedby 225

adjustingtheinitialtemperatureofthereactionchamber surfaces 226

(maximumtemperatureof393K) viaan electricalheatingsystem 227

around the reaction chamber. The compressed gas temperature, 228

T_C,wascalculatedfromtheinitialtemperature,T_i,initialpressure, 229

P_i, reactant composition, and the experimentally measured com- 230

pressed gas pressure, P_C. Forthis calculation, the adiabatic com- 231

pression/expansionroutine inGaseq[27],whichusesthetemper- 232

aturedependenceofthespecificheatratioofthetest mixture,

γ

^,

233

accordingtotheequationEq.(1),wasemployed.

234

ln



_p

C

pi



=

_TC

Ti

γ γ

− 1 dT

T (1)

The signal from a Kistler603B piezoelectronic pressure trans- 235

ducer, installed in the combustion chamber, monitored the pres- 236

sureduringeach experimentandwasrecordedusingadigitalos- 237

cilloscope.Theignitiondelaytimewasdefinedasthetimeinterval 238

fromthe end of compression (first local maximum on the pres- 239

suretrace)to themaximum rateofpressure rise dueto ignition 240

(Fig.3). Anuncertaintyof± 15% isstimated fortheRCM ignition 241

delaytimemeasurements.

242

Non-reactive experiments were carried out under the same 243

conditions asthe corresponding reactive case in order to obtain 244

pressure-timehistorieswhichareconvertedtovolume-timehisto- 245

ries(using theisentropicrelationship betweenpressureandden- 246

sity) to be used in the simulations of the reactive mixtures. In 247

this way, facility effects including reaction during compression 248

andheat lossare taken intoaccount [32].The non-reactivemix- 249

tures were prepared by substituting O₂ in the reactive mixture 250

forthe diluent gas being used. Due to the unreactive behaviour 251

of DMC, the diluent gas used in the RCM experiments was Ar, 252

whichhasalowerheatcapacitythanN₂.Inthisway,thespecific 253

heatratio(

γ

=Cp/Cv)ofthetestgasincreasesallowing toachieve 254 highercompressedtemperatures.Figure3presentsatypicalpres- 255 suretracemeasuredintheRCMwiththereactiveandnon-reactive 256 mixture.AsobservedinFig.3,onlypressureriseduetocompres- 257 sionofthegasisdetectedduringthenon-reactiveexperiments. 258

3. Modeling 259

CHEMKIN-PRO[33]wasusedtosimulateourmeasuredignition 260 delaytimes.Forthesimulations intheshocktubes,constantvol- 261 umeconditionswereassumed,whilefortheRCMvariablevolume- 262 timehistorieswereemployedtoincludefacilityeffects[32]. 263 As mentioned in the introduction, heretofore three sub- 264 mechanismshavebeenproposedintheliteraturetodescribeDMC 265 oxidation[9,16,18].Thefirstonewasdevelopedin2005byGlaude 266 et al. [9]. This sub-mechanism was added to a base mechanism 267 previouslydevelopedforDMMandDME[10,11].Thereactionrate 268 constants for reactions involving DMC were obtained by analo- 269 gies based on the reaction rate constants for other oxygenated 270 compoundsincludingdimethyl ether,formicacid, andmethylbu- 271 tanoate. The complete model (henceforth called the GlaudeDMC 272 model)consistsof102speciesand442reactions. 273 Adecadelater, Huetal.[16]developedanewsub-mechanism 274 for DMC oxidation which contains the same reaction classes as 275 thosecontainedinthesub-mechanismofGlaudeetal.[9](i.e.uni- 276 molecular decomposition, H-atom abstraction, ether-acid conver- 277 sion,H-atomabstractionofCH₃O(C=O)OHformedfromtheether- 278 acid conversion, and radical decomposition. There are a total of 279 24elementary reactions).Mostof therateconstantsof thesere- 280 actions are the same as in the DMC sub-mechanism of Glaude 281 et al. [9], while other rate constants were modified to satisfac- 282 torilypredict their ignitiondelaytime measurements. In particu- 283 lar,therateconstantsfortheCH₃OC(=O)O–CH3 andCH₃OC(=O)– 284 OCH₃ bond cleavagereactions, producingCH₃O(C=O)˙O+ ˙CH₃ and 285 CH₃O˙C=O+CH₃˙O, respectively, were modified by those recom- 286 mended by Dooley et al. [34] for the C₃H₇–C(=O)O–CH3 and 287 C₃H₇–C(=O)–OCH₃ bond cleavage reactions in methyl butanoate. 288 Moreover,therateconstantsoftheH-atomabstractionfromDMC 289 by O₂, ˙H andÖ atoms,and ˙C2H₃, ˙C2H₅, ˙CH3, CH₃˙O, CH₃˙O2, H˙O2 290 radicals were assumed to be identical to those for methyl bu- 291 tanoate. As for the base mechanism, Hu used the AramcoMech 292 1.3 mechanism [17]. The complete model (henceforth called the 293 HuDMCmodel)consistsof275speciesand1586reactions. 294 One year later (2016), Sun et al. [18] developed a new sub- 295 mechanism for DMC conversion including newtheoretical deter- 296 minationsandupdated ratesfromliterature and analogies.Com- 297 paring this most recent sub-mechanism with those of Glaude 298 et al. [9] and Hu et al. [16], new reactions were added (e.g. 299 the unimolecular decomposition reactions CH₃O(C=O)OCH₃  300 CH₃O˙CHO+CH₂O and CH₃O(C=O)OCH₃  CH3O(C=O)OCH+H₂; 301 the hydrogen abstraction reaction CH₃O(C=O)OCH₃+H˙CO  ³⁰² CH₃O(C=O)O˙CH2+CH₂O; and additionally the recombination re- 303 action of ˙CH3 and CH₃OC(=O)O˙CH2 radicals), while other reac- 304 tions were omitted (those involved in the ether-acid conversion 305 andH-atom abstractionfromCH₃O(C=O)OH).In total,Sun’s sub- 306 mechanismfortheDMCconversionincorporates23elementaryre- 307 actions.TheAramcoMech1.3[17]wasalsousedasthebasemech- 308 anism.Thecompletemodel(henceforthcalledtheSunDMCmodel) 309 consistsof257speciesand1563reactions. 310 Mostrecently,Alzueta etal.[19]developedamodelbyadding 311 the DMCsub-mechanism fromGlaude etal.[9] to abase mech- 312 anism describedand updated by the authors [20–24]. The com- 313 plete model (henceforth called the AlzuetaDMC model) consists 314 of284speciesand1206reactions. Alzuetaetal.[19],highlighted 315 theimportanceofthethermodynamicdataofthespeciesinvolved 316 intheDMCsub-mechanismdescribingDMCoxidation,specifically 317

Table 3

Enthalpy of formation of DMC and its associated species used in the GlaudeDMC, HuDMC, SunDMC and AlzuetaDMC models.

H °f298K (kcal/mol) Species

GlaudeDMC and HuDMC (CBS-Q method)

SunDMC (Group additivity method using THERM [35] software)

AlzuetaDMC (Group additivity method using THERM [35] software)

CH 3 O(C = O)OCH 3 –136.06 –135.85 –136.05

CH 3 O(C = O)O ˙CH 2 –88.10 –86.843 –91.59

CH 3 O(C = O)OH –140.93 – –141.77

CH 3 O(C = O) ˙O –82.29 –83.85 –89.90

˙CH 2 O(C = O)OH –93.64 – –97.39

Table 4

Rate coefficients in form of k = AT ⁿ exp(-E/RT) for reaction CH 3 O(C = O)OCH 3  CH 3 O(C = O) ˙O + CH 3 (units: cm ³ /mol/s/cal) in the GlaudeDMC, HuDMC, SunDMC and AlzuetaDMC models.

Model Reaction A n E

GlaudeDMC AlzuetaDMC CH 3 O(C = O) ˙O + ˙CH ³  CH ³ O(C = O)OCH ^3a 3. 00 × 10 ¹³ 0.00 0.00 HuDMC ^b CH 3 O(C = O)OCH 3 ( + M)  CH 3 O(C = O) ˙O + ˙CH 3 ( + M) ^b 2. 55 × 10 ²³ –1.99 8. 81 × 10 ⁴

low/ 1. 74 × 10 ⁷³ –1. 60 × 10 ¹ 8. 53 × 10 ⁴ / troe / 2. 18 × 10 ^–1 1.0 6. 37 × 10 ³ 8. 21 × 10 ⁹ / SunDMC CH 3 O(C = O)OCH 3  CH 3 O(C = O) ˙O + ˙CH 3c plog/ 0.04 2. 86 × 10 ⁷⁵ –17.58 112,569.

plog / 0.1 7. 95 × 10 ⁷² –16.71 112,126.

plog / 0.5 1. 13 × 10 ⁶⁷ –14.82 110,484.

plog / 1 1. 29 × 10 ⁶⁴ –13.89 109,488.

plog / 10 9. 42 × 10 ⁵⁹ –12.43 110,018.

plog / 100 3. 51 × 10 ⁵² –10.08 108,107.

plog/ 10 0 0 8. 70 × 10 ²⁶ –3.51 79,326.

a Assumed the same rate as high pressure rate for CH 3 ˙O + ˙CH 3 → CH 3 OCH 3[10] .

b Adopted from the decomposition of methyl butanoate in the work of Dooley et al . [34] .

c Calculated using the master equation code-PAPER [36] .

the thermodynamic data of the CH₃O(C=O)˙O radical species.The 318

enthalpy offormation ofDMC and its associated speciesin each 319

model,withthecorrespondingmethodsfortheircalculations, are 320

providedinTable3. 321

The performance ofall four models in predictingthe ignition 322

delaytimesofthiswork, aswell asthe influenceofthe thermo- 323

dynamicdata of theCH₃O(C=O)˙Oradical species on theignition 324

delaytime calculatedbyeach modelwillbe discussedinthefol- 325

lowingsection.

326

4. Resultsanddiscussion 327

This section is divided into three parts. The first part shows 328

theperformance ofthefouraforementionedmodels(GlaudeDMC, 329

HuDMC,SunDMC, andAlzuetaDMC)insimulatingtheexperimen- 330

tal ignition delay times of this work. The second investigates 331

the influence of the thermodynamic of the CH₃O(C=O)˙O radical 332

speciesonthecalculationsoftheignitiondelaytimesforallfour 333

models, by changing the thermodynamic of this radical in each 334

modelby thethermodynamic ofthisradicalintheother models.

335

The third provides a chemical interpretation of the effect of the 336

DMCconcentration,pressureandequivalenceratioontheignition 337

delay timesfor DMC oxidation by performing rate ofproduction 338

(ROP)andsensitivityanalyses.

339

4.1. Performanceofthemodels 340

Examples of the performance of the GlaudeDMC, HuDMC, 341

SunDMC and AlzuetaDMC models to predict the measured igni- 342

tion delaytimesatlow(1.75% DMC,

ϕ

=1,P=2atm,mixture 5A 343

in Table 1) and high (

ϕ

=1, P=20 atm, mixtures 2B and 2C in 344

Table2)pressureareprovidedinFig.4. 345

Itcanbeseenthatatlowpressure(1.75%DMC,

ϕ

=1,P=2atm, 346

Fig. 4a) the GlaudeDMC model significantly overpredicts the ex- 347

perimentalignitiondelaytimesovertheentiretemperaturerange 348

studied,while theHuDMC,SunDMCandAlzuetaDMCmodels un- 349

derpredict them. Regarding themodeling athighpressure (

ϕ

= 1, 350

P=20atm,Fig.4b),theGlaudeDMCandAlzuetaDMCmodelsshow 351 the poorest predictions, especially at low temperatures(≤ 870K). 352 Ontheotherhand,theHuDMCmodelagreeswiththeignitionde- 353 lay times athigh temperatures (≥ 900K), while it showsa large 354 underprediction of the ignition delay times at low temperatures 355 (≤ 870K). The SunDMC model predicts the ignition delay times 356 over the entiretemperature rangestudied moderately well, with 357 onlyaslightunder-predictionatthehighesttemperaturesstudied 358 (≥ 1250K). The same behavior in predictions is observed for all 359 the remaining experimental data at low and high pressure (not 360

shown). 361

Overall, the SunDMC model best predicts the measured data 362 athighpressureover theentiretemperaturerangestudied.How- 363 ever,thegoodpredictionathighpressurebutnotatlowpressure 364 (largeunderprediction)indicates that thismodelshowsaweaker 365 pressure dependence compared to that observed in the experi- 366 ments. This could be attributedprincipally to the uncertainty in 367 therate coefficientsof the unimolecular decompositionreactions 368

ofthefuelintheSunDMCmodel. 369

4.2.EffectofthethermodynamicoftheCH3O(C=O)˙Oradicalspecies 370

onthesimulations 371

As discussed earlier/above, Alzueta et al. [19] found that the 372 thermodynamicdataoftheCH₃O(C=O)˙Oradicalspeciesgreatlyin- 373 fluencestheircalculationsofDMCoxidation.Thus,toobservethis 374 influenceonthecalculationoftheignitiondelaytimedataofthe 375 presentwork, thethermodynamicsofthisradicalwaschangedin 376 eachofthefourmodels(GlaudeDMC,HuDMC,SunDMC,andAlzue- 377 taDMC)andsimulations were run. Thischangein thermodynam- 378 ics only represented an effect on the simulation data using the 379 GlaudeDMCandAlzuetaDMC models (Figs. 5 and6,respectively) 380 throughoutthetemperaturerangestudiedatlowpressure,andat 381 hightemperatures(≥ 950 K)athighpressure,whilethesimulation 382 dataathighpressureandlowtemperatures(≤ 870K) wasnot af- 383

fected. 384

Fig. 4. Measured (symbols) ignition delay times in conjunction with calculations (lines) using GlaudeDMC, HuDMC, SunDMC and AlzuetaDMC models. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

The unimolecular DMC decomposition reaction 385

CH₃O(C=O)OCH₃  CH3O(C=O)˙O+CH₃ (R1), reported to be im- 386

portantfor thefuel consumption athigh temperatures[13,16,18], 387

and which rate constant is the same in both GlaudeDMC and 388

AlzuetaDMC models (Table 4), wasidentified to be the cause of 389

this event. When this reaction in GlaudeDMC and AlzuetaDMC 390

modelswasreplacedbythat intheHuDMCandSunDMCmodels, 391

the thermodynamics of the CH₃O(C=O)˙O radical species had no 392

anyeffectonthesimulationdata,ascanbeseeninFigs.7and8, 393

respectively.

394

Moreover, when reaction R1 in HuDMC and SunDMC models 395

wasreplaced by that in the GlaudeDMC/AlzuetaDMC model,the 396

thermodynamicsoftheCH₃O(C=O)˙Oradicalspeciesinfluencedthe 397

calculationsbythesetwomodels(Figs.9and10).AsinFigs.7and 398

8, thisinfluenceoccursall overthetemperaturerangestudied at 399

lowpressure,whileathighpressure,thereisnoteffectofthether- 400

modynamicoftheCH₃O(C=O)˙Oradicalspeciesonthecalculations 401

atlowtemperature,whichindicatesthatreactionR1doesnotcon- 402

trol the prediction of ignition delay times at low temperatures.

403

Alsonote that calculationswiththe HuDMCmodel usingthe re- 404

actionR1 intheSunDMCmodel (Fig.9), andviceversa (Fig.10), 405

werenot affectedwhen thethermodynamicsoftheCH₃O(C=O)˙O 406

radicalspecieswaschanged.

407

Figures 4–10 show that simulations with the SunDMC model 408

that incorporate reaction R1 from the HuDMC model (Fig. 10) 409

(modifiedSunDMCmodel,henceforthcalledSunDMC_mod)agree 410

betterwiththeexperimental dataat1.75% DMC,

ϕ

=1,P=2atm 411

Fig. 5. Effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species on the cal- culations of the ignition delay time by the GlaudeDMC model. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

(Fig.10a),and

ϕ

=1,P=20atm(Fig.10b),thantheoriginal(Fig.4) 412

andthemodified(Figs.5–9)models. 413

Besides reaction R1, the reaction CH₃O(C=O)OCH₃  414 CH₃OCH₃+CO₂ (R2) has been reported [13,18] to be a domi- 415 nant decomposition channel, and thus competes with reaction 416 R1. Thereby, the ignition is also influenced by the k₁/(k₁+k₂) 417 branching ratio. The rate constant, k₁ in Hu et al. [16] was 418 adoptedfromthedecompositionofmethylbutanoateinthework 419 of Dooley et al. [34], and is in reasonable agreement with the 420 reaction rate measured directly by Peukert et al. [13]. On the 421 other hand,k₂ usedby Sunetal.[18]wasdetermined bymaster 422 equationanalysisandisalsoinagreementwiththat measuredby 423 Peukertetal.[13],beingimprovedinrelationtothatused byHu 424 etal. [16]which wastaken fromthe estimation of Glaudeet al. 425

[9]basedonreactivityanalogy. 426

Calculations for the remaining experimental conditions stud- 427 ied (1.75% DMC:

ϕ

=0.5 and 2.0; 0.75% DMC:

ϕ

=0.5, 1.0 and 428 2.0; P=20 atm:

ϕ

=0.5 and 2.0; P=40 atm:

ϕ

=0.5, 1.0 and 429 2.0) were performed using the SunDMC_mod model.This model 430 predicts our measured ignition delay times at both low and 431 high pressures and at all of the temperatures and equivalence 432 ratios studied. As mentioned earlier, the SunDMC_mod model 433 uses AramcoMech1.3 asthe basemechanism, and replacingthis 434 withAramcoMech 2.0[37], i.e., the sub-mechanism forthe DMC 435 conversion in the SunDMC_mod model has been added to the 436 AramcoMech2.0mechanism(thisfinalmodelwillbecalledAram- 437 coMech2.0+SunDMC_mod), had no effect on our ignition delay 438 timepredictions.Thethermodynamicdatausedcorrespondtothe 439 samesourcesasthereactions,i.e.,allthethermodynamicdataare 440 fromAramcoMech2.0exceptthethermodynamicsofDMCandits 441

Fig. 6. Effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species on the cal- culations of the ignition delay time by the AlzuetaDMC model. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 2 atm (mixtures 2B and 2C in Table 2 ).

associated species which are fromSunDMC model. This updated 442

mechanismisprovidedasSupplementaryMaterial.

443

Inthefollowingsection,theAramcoMech2.0+SunDMC_modis 4 4 4

usedtoperform thesimulations(which resultsarethesamethat 445

usingtheSunDMC_modmodel),aswell asthechemical interpre- 446

tationoftheeffectoftheDMCconcentration,pressureandequiv- 447

alenceratioonDMCignitiondelaytimes.

448

TheperformanceoftheAramcoMech2.0+SunDMC_modmodel 449

inpredictingtheconcentrationprofilesofthespeciesmeasuredin 450

the flame studies of Sinhaand Thomson [8] andSun etal. [18], 451

aswellasintheflowreactorsofSunetal.[18]andAlzuetaetal.

452

[19]areprovidedinFigs.S1–S5oftheSupplementarymaterial.

453

4.3. Effectoftheparametersontheignitiondelaytime.Chemical 454

interpretations 455

4.3.1. EffectoftheDMCconcentration 456

Figure 11shows the effect of DMC concentration under fuel- 457

lean (

ϕ

=0.5), stoichiometric (

ϕ

=1), and fuel-rich (

ϕ

=2) condi- 458

tions (mixtures1A–6AinTable1). Similarto otherhydrocarbons, 459

andasreportedby Huetal.[16],an increase inDMCconcentra- 460

tion increasesreactivity,i.e., it decreasesignition delaytimes,for 461

alloftheequivalenceratiosstudied.

462

TofurtherstudytheeffectoftheDMCconcentrationontheig- 463

nition delaytime, a reactionpathwayanalysis (ROP) anda brute 464

forcesensitivityanalysiswereperformedfor0.75%and1.75%DMC, 465

at

ϕ

=1,P=2 atmat T=1350K.Sensitivity coefficients(S_i) were 466

calculated using an automated code developed in-houseand are 467

expressedaccordingtoEq.(2). 468

S_i=log₁₀

( τ

k×2/

τ

k/2

)

log₁₀

(

²/0.5

)

⁽²⁾

Fig. 7. Effect of reaction R1 on the calculations of the ignition delay time by the GlaudeDMC model, and effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species with the change of reaction R1. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

where

τ

kx2 isthe simulatedignitiondelaytimewhen aratecon- 469 stantisincreasedbyafactorof2and

τ

k/2isthatpredictedwhen 470 arateconstantisdecreasedbyafactorof2.Therefore,anegative 471 sensitivitycoefficientdenotes arateconstant thatpromotes reac- 472 tivity(decreasing ignition delaytime by increasing the ratecon- 473

stant),andviceversa. 474

The reaction pathwayanalysis depicted inFig. 12,which was 475 carriedoutatthetime of20% DMCconversion,indicatesthat the 476 majorpathsassociatedwithDMCconsumption aretwoDMCuni- 477 molecular decompositions reactions and the H-atom abstraction 478 fromDMC,principallyby ˙Hatomsand ˙OHradicals. 479 ThemostimportantpathwayforDMCconsumption isits uni- 480 molecular decomposition to produce dimethyl ether (CH₃OCH₃) 481 andCO₂ (reaction R2).35%and32.4% ofthe fueldecomposesby 482 thisrouteforthe0.75%and1.75%DMCmixtures,respectively. 483 TheCH₃OCH₃soformedcanbelargelyconsumedviatwomain 484 pathways:(a) H-atom abstractionby ˙H atoms(31.1%,25.8%), and 485

˙OH (25.2%, 30.6%) and ˙CH₃ (15.9%, 19.2%) radicals(with percent- 486 ages in brackets corresponding to mixture of 0.75% and 1.75% 487 DMC, respectively) to lead, in a further step, to the formation 488 of formaldehyde and methyl radicals; or (b) decomposition to 489 give ˙CH3 and methoxy (CH₃˙O) radicals (23.4 %, 19.7%) (reaction 490 R3) which then principally decompose via reaction R4 to give 491 formaldehydeand ˙Hatoms(93.6%,86.5%). 492

CH3OCH3 (+M) ˙CH3+CH3˙O(+M) (R3) CH₃˙O(+M) CH2O+ ˙H(+M) (R4)

The other important unimolecular decomposition reaction for 493 DMC consumption (although, based on our reaction pathway 494 analysis, it is the less important route) is reaction R1, which 495

Fig. 8. Effect of reaction R1 on the calculations of the ignition delay time by the AlzuetaDMC model, and effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species with the change of reaction R1. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

consumes23.2%and21.9%ofthefuelforthemixtureswith0.75%

496

and1.75%DMC,respectively. TheCH₃O(C=O)˙O radicalformed to- 497

tallydecomposestoCO₂ andCH₃˙Oradicals whichmainlydecom- 498

posestoformaldehydeand ˙HatomsviareactionR4.

499

H-atom abstraction fromDMC,principally by ˙H atoms(24.4%, 500

22.7%),toproduceDMCradicals(CH₃O(C=O)O˙CH2)andmolecular 501

hydrogen(reaction R5),isthesecond-mostimportantreactionfor 502

theconsumptionofDMCundertheconditionsanalysed.

503

CH3O(C=O)OCH3+ ˙H CH3O(C=O)O˙CH₂+H2 (R5) The CH₃O(C=O)O˙CH2 radicals formed can decompose to 504

formaldehyde andmethoxy formyl radicals (CH₃O˙C=O) (reaction 505

R6),whichinturns decomposeto ˙CH₃ radicals andCO₂ (reaction 506

R7).

507

CH3O(C=O)O˙CH₂  CH3O˙C=O+CH2O (R6)

CH3O˙C=O ˙CH3+CO2 (R7)

The sensitivity analysis provided in Fig. 13 indicates that the 508

mostimportant reactions promoting reactivity are the ˙H+O₂  509

Ö+ ˙OH (R8) chain-branching reaction and the fuel-specific reac- 510

tionR1.However, thechain-branchingreactionbecomeslesssen- 511

sitivewithincreasingfuelconcentration,whilethespecific-fuelre- 512

actionbecomesmoresensitivewithincreasingfuelconcentration.

513

Thehighdependenceofthe1.75%DMCmixtureonreactionR1can 514

explain the decrease inignition delay times withincreasing fuel 515

concentration,sincethehighertheDMCconcentration,thehigher 516

thereactionrateofreactionR1leadingtoshorter(faster)ignition 517

delaytimes.

518

Fig. 9. Effect of reaction R1 on the calculations of the ignition delay time by the HuDMC model, and effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species with the change of reaction R1. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

Fig. 10. Effect of reaction R1 on the calculations of the ignition delay time by the SunDMC model, and effect of the thermodynamic of the CH 3 O(C = O) ˙O radical species with the change of reaction R1. a) 1.75% DMC, ϕ= 1, P = 2 atm (mixture 5A in Table 1 ), and b) ϕ= 1, P = 20 atm (mixtures 2B and 2C in Table 2 ).

Fig. 11. Effect of DMC concentration on ignition delay times at P = 2 atm and ϕ= 0.5 (mixtures 1A and 4A in Table 1 ), 1.0 (mixtures 2A and 5A in Table 1 ), and 2.0 (mixtures 3A and 6A in table 1 ). Symbols: experimental data. Lines: Aram- coMech2.0 + SunDMC_mod mode ling.

Reaction R1leadstotheformationof ˙CH3 radicals,whichpro- 519

mote reactivity, mainly forthe 1.75% DMC mixture, dueto reac- 520

tionsR9andR10.

521

˙CH₃+H˙O₂  CH3˙O+ ˙OH (R9) CH2O+˙CH₃ H˙CO+CH4 (R10) Methoxy(CH₃˙O) andformyl(H˙CO) radicalsfurther decompose 522

byreactionsR4andR11,respectively,givingreactive ˙Hatomsand 523

thenleadingtotheincreaseofthereactionrateofthehighsensi- 524

tivepromoterreactionR8.

525

H˙CO+M ˙H+CO+M (R11)

Fig. 12. Reaction pathway analysis for 0.75% DMC (italic font) and 1.75% DMC (bold font), ϕ= 1.0, P = 2 atm, T = 1350 K. Carried out at the time of 20% fuel conversion.

Fig. 13. Sensitivity analysis for 0.75% DMC and 1.75% DMC, ϕ= 1.0, P = 2 atm, T = 1350 K.

Other important promoting reaction, mainly for the mixture 526 with0.75%DMC,isthedimethyletherdecompositionreactionR3. 527 Thisreactionpromotesreactivitybecausebesidesforming ˙CH3rad- 528 icals, italso produces reactive ˙Hatoms via thedecomposition of 529

CH₃˙Oradicals(reactionR4). 530

H-atomabstractionfromDMCby ˙Hatoms(reaction R5),isthe 531 most inhibiting reaction as it forms a stable hydrogen molecule 532 froma very reactive hydrogenatom, and thisreaction alsocom- 533 petes with the most important chain-branching reaction R8, by 534 consuming approximately 47% of the total concentration of ˙H 535 atoms,resultinginareducedreactivity. 536 The other reactions inhibiting reactivity are termination reac- 537 tions which involveconsumption of ˙CH₃ radicals (mainlyfor the 538 mixtureof1.75% DMC) byreactions R12andR13, andvery reac- 539 tive ˙H atoms(mainly forthemixture of0.75% DM) viareactions 540

R14andR15,toproducestablespecies. 541

˙CH₃+H˙O ₂ CH4+O₂ (R12)

˙CH₃+˙CH₃ (+M) C2H₆(+M) (R13)

Fig. 14. Effect of the pressure on ignition delay times of DMC for ϕ= 0.5 (mix- tures 1B, 4B, 1C and 4C in Table 2 ), 1.0 (mixtures 2B, 5B and 2C in Table 2 ) and 2.0 (mixtures 3B and 6B in Table 2 ). Symbols: experimental data. Lines: Aram- coMech2.0 + SunDMC_mod mode ling.

H˙O₂+ ˙H H2+O2 (R14)

˙CH₃+ ˙H(+M) CH4 (+M) (R15) 4.3.2. Effectofthepressure

542

Figure 14 showsthe effect ofpressure on DMCignition delay 543

timesatallthreeequivalenceratiosstudied(mixtures1B–6Band, 544

1C,2Cand4CinTable2).Noignitionwasobservedwiththefuel- 545

richmixturecompressedat20atmintheRCM(3CinTable2)and, 546

asmentioned in Section 2, experiments with stoichiometric and 547

fuel-richmixtures compressedto40atmcould notbe performed 548

intheRCM.

549

The ignition delaytime decreases withincreasing pressure at 550

allequivalence ratios,witha morepronounced pressure effectat 551

Fig. 15. Reaction pathway analysis at a) T = 860 K and b) T = 1350 K for P = 20 (italic font) and 40 atm (bold font), ϕ= 0.5. Carried out at 20% fuel consumption.

temperatures below approximately 1250K. This increase in reac- 552 tivitywith increasing pressureis expectedasathigher pressures 553 theabsoluteconcentration ofreactantsincreases,therebyincreas- 554

ingthereactionrate. 555

Rateofproduction/ROPandsensitivityanalyseswereperformed 556 at860Kand1350Kfor20and40atm,

ϕ

=0.5,andareshownin 557 Figs.15and16,respectively.Theanalyseswerecarriedoutatlow 558 (860K) and high(1350K) temperatureto account forthe largest 559 differenceinignitiondelaytimesobservedattemperaturesbelow 560

1250K. 561

As shown in Fig. 15, at a given temperature, the main reac- 562 tion pathways forDMC consumption are the same atboth pres- 563 sures. However, there are differences in these at low and high 564 temperatures. At highpressures andtemperatures (Fig. 15b), the 565 reaction flux resembles that at low pressures and high temper- 566 atures (Fig. 12), withthe difference being that, at high pressure 567 the unimolecular fuel decomposition reaction R1 is not impor- 568 tantwhileH-atomabstractionreactionsareimportant.H-atomab- 569 stractionfromDMC,mostlyby ˙OHradicals,isthedominantpath- 570 way consuming DMC at both pressures, being slightly higher at 571 40atmcomparedto20atm(55.1%and47.6%,respectively).Onthe 572 otherhand,athighpressuresandlowtemperatures(Fig.15a),fuel 573 consumptionis controlledby H-atomabstractionreactions by ˙OH 574

Fig. 16. Sensitivity analysis at a) T = 860 K and b) T = 1350 K for P = 20 and 40 atm, ϕ= 0.5.

radicals, withthe percentageof DMC consumption being almost 575

thesameatbothpressures(84.1%and84.3%at20atmand40atm, 576

respectively).

577

Figure 16 indicates that, at a given temperature, the reac- 578

tions controlling reactivityarethe sameatboth pressures,which 579

suggeststhat thedecrease inignition delaytime with increasing 580

pressure,observedinFig.14,islargely dueto theincreaseinab- 581

soluteconcentrationofreactants(DMCandO₂)athigherpressure, 582

ratherthantoachangeinthecontrollingchemistry.However,the 583

reactions controlling the reactivity atlow and hightemperatures 584

arecompletelydifferent.

585

At low temperatures (860K, Fig. 16a), there are three fuel- 586

specific reactions (R16,R17,andR18)that stronglypromotereac- 587

tivity.

588

CH3O(C=O)OCH3+CH3˙O₂  CH3O(C=O)O˙CH₂+CH3O2H (R16) CH3O(C=O)OCH3+H˙O₂  CH3O(C=O)O˙CH₂+H2O2 (R17) CH₃O(C=O)OCH₃+˙OH CH3O(C=O)O˙CH₂+H₂O (R18)

The CH₃O(C=O)O˙CH2 radicals formed from these three re- 589 actions can further decompose to produce formaldehyde and 590 CH₃O˙C=Oradicals(reaction R6),whichinturndecomposetopro- 591 duce ˙CH3 radicals andCO₂ (reaction R7). Moreover, the CH₃O₂H 592 and hydrogen peroxide (H₂O₂) molecules formed, decompose to 593 formoneandtwo ˙OHradicalsthroughreactionsR19andR20,fur- 594

therpromotingreactivity. 595

CH₃O₂H CH3˙O+ ˙OH (R19)

H₂O₂ (+M) ˙OH+ ˙OH(+M) (R20)

Thus,for eachmolecule of fuelthat reactswithmethylperoxy 596 (CH₃˙O2)andhydroperoxyl(H˙O2)radicals,oneandtwo ˙OHradicals 597 areformed,respectively. Thisgreatlypromotes reactivitybecause, 598 aspreviously discussed in relation to Fig. 15, the main path for 599 DMCconsumptionathighpressures isH-atomabstractionby ˙OH 600 radicals,which inturn alsopromotes reactivityathighpressures 601

andlowtemperatures(Fig.16a). 602

Fig. 17. Effect of equivalence ratio on ignition delay times of a) 0.75% DMC and b) 1.75% DMC at P = 2 atm (Mixtures 1A –6A in Table 1 ). Symbols: experimental data.

Lines: AramcoMech2.0 + SunDMC_mod mode ling.

Another reaction that promotes the formation of ˙OH radicals 603

and hence increases reactivity, is the reaction of formaldehyde 604

with the CH₃˙O₂ radicals to produce H˙CO radicals and CH₃O₂H 605

molecules.Thispromotesreactivitybecauseitleadstotheforma- 606

tionof ˙OHradicalsthrough (1)thedecompositionoftheCH₃O₂H 607

molecules(reaction R19),and(2)thereactionoftheH˙COradicals 608

withmolecularoxygenbythefollowingreactionsequence:

609

H˙CO+O2  CO+H˙O₂ (R21)

H˙O2+H˙O2 H2O2+O2 (R22) Thereafter the H₂O₂ molecules decompose via reaction R20, 610

leadingtotheformationoftwo ˙OHradicals.

611

AsitisobservedinFig.16a,reactionR22inhibitsreactivitybe- 612

causethisterminationreactioncompeteswithH-atomabstraction 613

fromDMC (reactionR17). IfH˙O₂ radicals reactwithDMC instead 614

ofreacting with each other, two H₂O₂ molecules can be formed 615

(throughreactionR17),ultimatelyleadingtotheformationoffour 616

˙OHradicals(throughreactionR20),therebyincreasingreactivity.

617

Theotherreactionthatinhibitsreactivityistheterminationre- 618

actionof CH₃˙O2 with methylradicals to produce two CH₃˙O rad- 619

icals. In this way, CH₃˙O₂ radicals are consumed to produce sta- 620

blespeciesinsteadofreactingwithDMC(byreactionR16)tofur- 621

therformthe ˙OHradicals through thedecomposition ofCH₃O₂H 622

molecules(reactionR19).

623

Fig. 18. Effect of equivalence ratio on ignition delay time of DMC at P = 20 and 40 atm (Mixtures 1B – 6B and 1C, 2C and 4C in Table 2 ). Symbols: experimental data. Lines: AramcoMech2.0 + SunDMC_mod mode ling.

Fig. 19. Sensitivity analysis for ϕ= 0.5, 1.0, and 2.0, 1.75% DMC, P = 2 atm at T = 1350 K.

Thesensitivitycoefficientsofthesetwoinhibitingreactionsin- 624 crease with decreasing pressure, indicating that these two reac- 625 tionsaremoredominantatlowerpressures. 626 At hightemperatures (1350K,Fig. 16b), mostof the reactions 627 promotingandinhibiting reactivityaresimilar tothose discussed 628 above inFig.13.Reaction R1leadsto theformation of ˙CH3 radi- 629 calswhichpromotesreactivity,sincetheyreactwithhydroperoxyl 630