Analysis of acetic acid gas phase reactivity: Rate constant estimation and kinetic simulations

Proceedings of the Combustion Institute 37 (2019) 539–546

www.elsevier.com/locate/proci

Analysis

of

acetic

acid

gas

phase

reactivity:

Rate

constant

estimation

and

kinetic

simulations

Carlo Cavallotti

, Matteo Pelucchi

, Alessio Frassoldati

DepartmentofChemistry,Materials,andChemicalEngineering,PolitecnicodiMilano,viaMancinelli7,20131Milan, Italy

Received30November2017;accepted18June2018 Availableonline5July2018

Abstract

Thegasphasereactivityofaceticacidwasinvestigatedcombiningfirstprinciplecalculationswithkinetic simulations.Rateconstantsfortheunimoleculardecompositionofaceticacidweredeterminedintegrating the1DmasterequationoveraPotentialEnergySurface(PES) investigatedattheM06-2X/aug-cc-pVTZ level.EnergieswerecomputedattheCCSD(T)/aug-cc-pVTZlevelusingabasissetsizecorrectionfactor de-terminedattheDF-MP2/aug-cc-pVQZlevel.Threedecompositionchannelswereconsidered:CO2+CH4,

CH2CO+H2O,andCH3 +COOH.Rateconstantswerecomputedinthe700–2100Kand0.1–100atm

temperatureandpressureranges.Thesimulationsshowthatthereactionisinfalloffabove1200Kat pres-suressmallerthan10atm.Successively,thePESsforaceticacidH-abstractionbyH,OH,OOH,O2,and

CH3wereinvestigatedatthesameleveloftheory.Rateconstantswerecomputedaccountingexplicitlyfor

theformationofentranceandexitvanderWaalswellsandtheircollisionalstabilization.Energybarriers weredeterminedattheCASPT2levelforH-abstractionbyOHoftheacidicH,sinceithasastrong mul-tireferencecharacter.Thecalculatedrateconstantisingoodagreementwithexperimentsandsupportsthe experimentalfindingthatatlowtemperaturesitispressuredependent.Thecalculatedrateconstantswere usedtoupdatethePOLIMIkineticmodelandtosimulatethepyrolysisandcombustionof aceticacid.It wasfoundthataceticaciddecompositionandtheformationof itsdirectdecompositionproductscanbe reasonablypredicted.Theformationofsecondaryproducts,suchasH2andC2hydrocarbons,is

underpre-dicted.Thissuggeststhatreactionroutesnotincorporatedinthemodelmaybeactive.Somehypothesesare formulatedonwhichthesemaybe.

© 2018TheAuthors.PublishedbyElsevierInc.onbehalfofTheCombustionInstitute.

ThisisanopenaccessarticleundertheCCBYlicense.(http://creativecommons.org/licenses/by/4.0/ )

Keywords: Aceticacid;Abinitio;Masterequation;H-abstraction;Rateconstants

∗_{Correspondingauthor.}

E-mailaddress:carlo.cavallotti@polimi.it (C. Caval-lotti).

1. Introduction

Acetic acid is found in significant concentra-tionsinthetroposphereanditisknowntoaffect considerablytheenvironmentalchemistry.Its ori-ginisbothbiogenicandanthropogenic.Biomass

https://doi.org/10.1016/j.proci.2018.06.137

1540-7489© 2018TheAuthors.PublishedbyElsevierInc.onbehalfofTheCombustionInstitute.Thisisanopenaccess articleundertheCCBYlicense.(http://creativecommons.org/licenses/by/4.0/ )

combustionandvehicleemissionsarebelievedto beamongthemostimportantsourcesthatcanbe directlyrelatedtohumanactivities.Moreover, oxy-genated species are very abundant in the tar re-leasedfrombiomass pyrolysis.Acetic acidisthe majoracidiccomponent of bio-oilsderivedfrom biomass fastpyrolysis [1] . These renewable fuels areacidicliquids(pH2–3),largelydifferingfrom petroleum fuelswith regard to both their physi-calpropertiesandchemicalcomposition.The im-portanceofaceticacidonthereactivityof pyroly-sisproductsofcellulosewasrecentlydiscussedby Fukutomeetal.[2] .Themodelingofeach individ-ualconstituentofthebio-oilsisnotfeasible, there-fore thesecomplex mixtures arecharacterized in termsofalimitednumberofrepresentative chem-icalcomponents(surrogatemixture).Aceticacidis thereferencespeciesusedtorepresentcarboxylic acidsinbio-oilsurrogates[3] .Therefore,aproper knowledgeaboutitspyrolysisandcombustion be-haviorisnecessarytocharacterizethecombustion ofbiomasspyrolysisoils.Tocontroltheformation andthesuccessivereactivityofaceticaciditis nec-essarytounderstanditsformationand decomposi-tionpathways.Inthepresentworkwefocusonthe decompositionpathwaysofaceticacidand there-forewelimittheliteratureanalysistothisaspect.

Acetic acid can either decompose directly through a unimolecular process or through H-abstractionreactions.Theunimolecular decompo-sitionof aceticacidhasbeenthesubjectofboth theoretical and experimental studies [4–13] . The firstkineticstudywasperformedin1949by Bam-fordandDewarusingaflowmethodinaquartz tubeinthe773–1173Ktemperaturerange[4] .They foundthatthemaindecompositionreactionsare: CH3COOH→CH4+CO2(R1)

CH3COOH→CH2CO+H2O(R2)

SuccessivelyBlakeandJacksoninvestigatedthe samereactionatatmosphericpressureinthe733– 868 K temperature range [5] .In both casesrate constants were determined fitting the evolution profilesof theproducts,thusnot accounting for secondarychemistry.Thefirststudyincombustion relevantconditions(1300–1950K)wasperformed byMackieandDoolanusingasinglepulseshock tube[6] .Theexperimentalresultswereinterpreted usingakineticmodel,whichwasusedtofitR1and R2rate constants.A similarbranching of about 0.5wasfoundbetweenthetworeactionchannels. Two successive experimental studies by Saito et al. [7] and Butkoskaya et al. [8] found kR2/kR1

branchingratiosof1and2,respectively.Recently, Elwardanyetal.[9] studiedthesamesystemusinga shocktubeinthe1230–1821Ktemperaturerange. ThekR2/kR1branchingratiowasabout2,whilerate

constantswereaboutoneorderofmagnitudefaster thanthoseofMackieetal.Severaltheoretical

stud-ieshaveinvestigatedthedecompositionmechanism ofaceticacid.Thepotentialenergysurface(PES) wasfirststudiedbyNguyenetal.[10] andDuan andPage[11] ,andmorerecentlyby Clarket al.

[12] andSunetal.[13] .Allthesestudiesagreeonthe PESqualitatively,thoughthereissomedifference inthekR2/kR1 branching, predictedtobeabout2

intheearlierstudiesandabout1byClarketal.

The second mechanism of decomposition of

aceticacid isthrough bimolecularH-abstraction reactions.WhileH-abstractionfromOHhasbeen thesubjectof manystudiesintheliterature,both experimental[14–18] andtheoretical[17–19] ,only Mendesetal.[19] havesystematicallyinvestigated H-abstractionbyotherradicals,suchasCH3,H,

OOH,andO,whichcansignificantlycontributeto aceticacidreactivity.H-abstractionbyO2hasnever

beentheoreticallyor experimentallyinvestigated. Inaddition,moststudiesonabstractionreactions byOHwereperformedatroomtemperature.There isthusstillsomeuncertaintyconcerningthese re-actionchannels,andinparticularforthe compli-catedreactionbetweenOHandaceticacid,which hasevidencedsomepeculiarpressuredependence atroomtemperature[16] .

The purpose of the present work is to re-examineaceticacidreactivitycombiningfirst prin-ciplesrateconstantestimationsandkinetic simula-tions.First,pressuredependentrateconstantsfor theunimoleculardecompositionofaceticacidwere determinedsolvingthemulti-wellmasterequation onaPESdeterminedfromhighlevelabinitio cal-culations.ThenrateconstantsforH-abstractionby H,OH,OOH,CH3,andO2 weredetermined.All

electronicstructure calculations andrate estima-tionswere performedatalevel of theorythatis higherthanthatpresentlyreportedintheliterature. Thecapabilityof theupdatedaceticacidmodule of thePOLIMIkineticmechanismtopredict py-rolysis,laminarburningvelocities,andspeciation inlowpressureflamesisthendiscussed.

2. Methodologies

2.1. Theoreticalcalculations

Theunimoleculardecompositionrateofacetic acidwas determinedperformingmasterequation (ME)simulationsonanabinitioPES.ThePESwas initiallyinvestigatedattheM06-2X/6-311+G(d,p) levelandthenrefinedattheM06-2X/aug-cc-pVTZ level, used also to computethe Hessian matrix. Energies were determined at the CCSD(T)/aug-cc-pVTZ level and corrected with the difference betweenDensity-Fitting(DF)MP2energies com-putedusingaug-cc-pVQZandaug-cc-pVTZbasis sets[20] .Torsionalmotionsweretreatedinthe1D hinderedrotorapproximationusinga M06-2X/6-311+G(d,p)rotationalPES.Torsionalvibrational frequencies were projectedout fromthe Hessian

as suggested by Sharma et al. [21] , so that no humaninterventionisnecessary.Quantum tunnel-ingwasaccountedforusingtheEckartmodel.In additiontothereactionchannelsreportedinthe lit-eraturetwohomolyticdecompositionrouteswere considered:Hlossfromthemethylgroupand de-compositiontomethylandCOOH.Thetransition state densityof states(DOS) forall thereaction channelswasdeterminedusingvariational transi-tionstatetheory.Forreactionsproceedingthrough saddlepointsvariationaleffectswereaccountedfor determiningtheDOSalongtheintrinsicreaction coordinate pathway. The homolytic decomposi-tionreactionenergiesandvibrationalfrequencies alongthereactionpathwaywerecomputedatthe multireference CASPT2 level with the breaking bondlengthsspacedby0.2A.˚ Calculationswere performedusinga(4e,4o)activespacethatincludes theCH3–COOHσ andσ∗orbitalsandtheπ and

π∗ _{orbitalsof the carboxyl group.The resulting}

energy profilewasrescaledtomatchthereaction energycomputedattheCCSD(T)level.ME calcu-lationswereperformedat0.1,1,10,and100atm in the700–2100K temperature range.The colli-sionalenergytransferwasdescribedusingasingle exponentialenergytransfermodelwitha temper-aturedependentEdownof 260×(T/300)0.85cm−1.

The calculated rate constants were fitted to the modified Arrhenius form for each considered pressure.Thesameapproach,madeexceptionfor CASPT2calculations,wasusedtoinvestigatethe decompositionkineticsofCH2COOH,whichcan

be produced either from the decomposition of aceticacidor,moreeasily,throughHabstraction fromtheaceticacidmethylgroup.Computational detailsarereportedassupplementalinformation.

TherateconstantsforH-abstractionreactions by OH, H, CH3,OOH, and O2 were computed

solving atwo wellsME whose stationary points werethereactantandproductvanderWaals(vdW) wellsandthetransitionstatebywhichtheyare con-nected.EntranceandexitchannelsfromthePES werecomputedusingphasespacetheory.The ad-vantage of this approach is thatit allows to set properly the energy of the reactants as a lower boundforthereactingfluxandcontextuallyto ac-countfortheimpactofthedepthofthevdWwells ontunneling.Contributionsofreactivefluxesfrom vdW wellsthatget collisionally stabilizedcanas wellbecomputedproperly,astheirformationrate andreactivityisincludedintheMEsolution. In addition,therateof formation of thevdW wells setsunupperlimittotherateconstant.For consis-tencywiththestudyoftheunimolecularreaction channels,geometries,frequencies,andenergiesof allstationarypointweredeterminedatthe M06-2X/aug-cc-pVTZ level, while energies were com-puted at the CCSD(T) level using the protocol describedabove.Since bothreactantsandsaddle pointscanhaveaconsiderablenumberofinternal torsionaldegreesof freedom(four inthecaseof

abstractionfromOOH),thesearch forthe mini-mumenergyconfigurationofthewellandthe sad-dle point was performedstarting froman initial guessgeometry,specifiedinz-matrixformat,whose torsionalangleswerethenmodifiedusingarandom valueinthe0–360rangetogenerateuncorrelated guessstructures.Upto15independent optimiza-tionswere performedtosearch fortheminimum energysaddlepointstructure.A2Dhinderedrotor modelwasusedtodescribeHabstractionbyOH fromthemethylgroup[22] .Inonecase,foracidicH abstractionfrommethylbyOH,energieswere com-putedattheCASPT2level,asCCSD(T)T1 diag-nosticsevidencedasignificantmultireference char-acter.Detailsoftheusedactivespacearereported intheresultssection.

AllM06-2Xcalculationswereperformedusing theG09suiteof codes[23] ,whileCCSD(T), DF-MP2, andCASPT2 calculationswere performed using Molpro 2010 [20] . ME input files were generated using a new code, EStokTP, designed toperformautomaticallytheinvestigationof the torsionalconformationspace,toprojecttorsional motionsfrom Hessians, anddetermine 1-D PES for rotors [24] . ME simulations were performed using MESS [22] . A summary of the adopted computationalproceduresisreportedinTableS1.

2.2. Developmentofaceticacidkineticmodeland kineticsimulations

Therateconstantscomputedinthisstudywere usedtoupdatetheaceticacidsub-mechanismin thePOLIMIkineticmodel.Movingtowarda uni-ficationofcoremechanisms,thePOLIMI mecha-nismwasrecentlyrevisedcouplingtheH2/O2 and

C1/C2from Metcalfe et al.[25] ,C3 fromBurke et al. [26] , and heavier fuels from Ranzi et al.

[27] .Thethermochemicalpropertieswereadopted, whenavailable,fromtheATcTdatabaseofRuscic orfromBurcat’sdatabase[28] .Thekinetic mech-anism,withthermodynamicandtransport proper-ties,isavailableasSupplementalMaterial(SM)to thispaper.

Allthekineticmodel simulationspresentedin thisworkhavebeencarriedoutusingthesolversfor idealreactorsandlaminar1Dflamesimplemented in the OpenSMOKE++ software [29] . Mixture-averagediffusioncoefficientswereusedintheflame simulations, including also thermal diffusion for thespeciestransportequations.Alargenumberof gridpointswereadoptedtoensuregrid-insensitive results(above300).

Fig.1.Potentialenergysurface(kcal/mol)forthe decom-positionofaceticacidtoproducts.Energiescomputedat theCCSD(T)/aug-cc-pVTZlevelandcorrectedforbasis sizeeffectswithDF-MP2/aug-cc-pVQZlevelenergies.

3. Resultsanddiscussions

3.1. Aceticacidunimoleculardecomposition: temperatureandpressuredependence

ThePESforaceticaciddecompositionisshown inFig. 1 .Itconsistsoftwowells,CH3COOHand

CH2C(OH)2, connected by one transition state,

andoffourdecompositionchannels.

The calculated energy barriers are in good agreementwith thosedeterminedby Clarket al.

[12] at the M06-2X/6-311+G(2df,p) level, from which theydiffer by lessthan 1 kcal/mol, made exception for isomerization to CH2C(OH)2, for

whichourbarrieris1.4kcal/molhigher.The dif-ference is more significantwith respect to Duan andPage[11] ,whocalculated,usingtheG2method onCASSCFstructures,asimilarenergybarrierof 71.8kcal/molfordecomposition toCH4 +CO2,

butamuchhigherenergybarrierfordecomposition toCH2CO+H2O bothfromW1(76.4kcal/mol)

andW2 (72.6 kcal/mol).Itis likely thatthe dif-ference withourdataisduetotherelatively low leveloftheoryatwhichstructuresweredetermined, whichdoesnot accountfordynamiccorrelation. HomolyticdecompositiontoCH3andCOOH,

in-cludedinthePESforthefirsttimeinthiswork,has areactionenergyof91.8kcal/mol,ingood agree-mentwiththeATcTvalueof92.5kcal/mol[28] .

Rate constants calculated at 1 atm assum-ingthat W2concentration isat apseudosteady state with respectto W1 are compared with lit-erature values in Fig. 2 for decomposition to CH4 +CO2 andinFig.S1fordecompositionto

CH2CO+H2O. Branchingratios toCH4 + CO2

andCH3+COOHarereportedinFig. 3 .

Fig.2. Comparisonbetweenrateconstantsforaceticacid decompositiontoCH4+CO2calculatedinthisworkand literaturevalues[6,9,12,13] .

Fig. 3. Comparison between branching ratios (kCH4+CO2/ktotal) for acetic acid decomposition to

CH4+CO2calculatedinthisworkandliteraturevalues (bothexperimentalandtheoreticaldata determinedat 1atm)[6,9,12,13] .

Therateconstantsvaluesherecalculatedare in-termediatebetweenthetwoexperimental measure-mentsofMackieandDoolan[6] andElwardanyet al.[9] ,thoughtheyaremoresimilartotheformer. Itshouldbenotedthatinbothexperimentalworks therateswerefittedusingakineticmechanism,so thatuncertaintiesinthemechanismarereflectedin theestimates.Theagreementwiththehighpressure predictionsof Clarket al.isquitegoodforboth theCH4 + CO2 andtheCH2CO+H2O reaction

channels.Ontheotherside,theagreementwiththe recentcalculationsofSunetal.[13] isgoodonly fortheCH4 + CO2 channel,whiletherates

cal-culatedbySun etal.fortheCH3 +COOH and

theCH2CO+H2Ochannelsareaboutafactorof

2.5higherand3slowerthanourrates,respectively. ThedifferenceforwhatconcernstheCH3+COOH

channelismostlikelydeterminedbythefactthat Sunetal.usedphasespacetheorytodeterminerate constantsofbarrierlesschannel,whichisalevelof theorylowerthantheoneusedinthepresent calcu-lationsanditiseffectivemostlyinprovidinghigher limitsforrateconstants,ratherthanaccurate esti-mates.

Thecomparison between literatureand calcu-latedbranchingratios(Fig. 3 )showsthatour pre-dictionsareinagreementwithMackieet al.and Clarketal.,forwhichweusedourcalculatedW2– W1pseudosteadystateconcentrationtocompute globalrates, whilethereisanotable discrepancy withElwardanyetal.andSunetal.Thehomolytic decompositionroutecontributionbecomes signifi-cantabove1000K.

Thepredictedpressuredependenceof therate of decomposition toCH4 + CO2 is reportedin

Fig.S2andshowsthatfalloff effectsstartbeing significantabove1200K.Rateconstantscalculated atdifferentpressuresandinterpolatedinthe mod-ifiedArrheniusformarereportedinTablesS2and S3, while the rate constants andPES calculated fortheunimoleculardecompositionofCH2COOH

arereportedinTablesS4andS5andFig.S3.On thebasisoftheleveloftheoryusedtocalculatethe rateconstantsandthecomparisonwithliterature dataweestimatethattheuncertainty of ourrate constantsisaboutfactorof2.

3.2. Abstractionrates

Rateconstants forH-abstractionsby OH, H, CH3,OOH,and3O2fromthemethylandcarbonyl

groupsof aceticacidwerecomputedasdescribed in the method section. The calculated rates are generallywithinafactorof2ofthosedetermined byMendes et al.[19] (see Fig. S5a-iforadirect comparison).The differenceis mostlikely deter-mined bythe higher level of theory used in the presentwork:MEcoupled tovariationalvs con-ventionaltransitionstatetheorytodeterminethe rates, energy barriers estimated at the CCSD(T) level using large basis sets, and hindered rotor treatment.Weestimatethattheuncertaintyofour rateconstantsisafactorof2orbetter.Twomain deviationsarefoundwithrespecttoMendesetal. estimates. Thefirst is forOOHabstraction from thecarboxylgroup,whereourrateissmallerthan Mendes etal.by afactor increasingfrom2to4 withtemperature.Afactorof1.5isdeterminedby variationaleffects,notaccountedforbyMendes.

The remaining discrepancy, as the computed

energy barriers are similar (23.3 kcal/mol in the

presentworkvs24.9kcal/mol,whichexplainsthe better agreementat lower temperatures), ismost likelyduetotheestimationofthehinderedrotors DOSandvibrationalfrequencies.Thisisa signif-icantlyendothermicreaction(25.7kcal/mol)with asubmergedbarrierandaproductliketransition state.Becauseof that,severallowvibrational fre-quenciesarefoundatthesaddlepoint,indicating thatthetransitionstateisratherloose.Themanual removalofthehinderedrotorrotationalfrequency fromthetransitionstatefrequenciesmaybe com-plicatedinsuchcasesasitmaybemixedwithother internalmotions.Theprojectiontechniqueweuse inthisworkremovessuchambiguity.

The second reaction for which a consider-abledifferencewithMendes etal. isfoundis H-abstractionbyOH of theacidic H.Thoughthis reaction has been thesubject of several theoret-ical investigation [17–19] , it is interesting to no-ticethatitwasneverexplicitlyreportedthatithas astrong multireference character. The CCSD(T) T1diagnostic we computed for the saddle point fortheradicalhydrogenabstraction(TSA)isquite high,0.142.Inaddition,wefoundthat,similarly towhatfoundforformicacid[30] ,thecarboxylH atom can be abstractedthrougha second mech-anism,which involvesaproton coupled electron transfermechanism.Wewereabletofindalsothis secondsaddlepoint(TSB),whichhasaswellahigh T1diagnosticof0.046.Thetwosaddlepoint struc-turesarecomparedintheSM.Toevaluatereliably energybarriersweperformedmultireference calcu-lationsattheCASPT2levelforthisspecificreaction channel.Themultireferencenatureofthisreaction isdeterminedbothbytheelectronicdegeneracyof theporbitalsoftheOHreactantsandbytwo reso-nancestatesoftheCH3COOproduct.Toaccount

properlyforthechangeintheelectronicstructure thattakesplacegoingfromreactantstoproducts theminimumactivespacemustberelativelylarge. Theactivespaceweusedconsists of15 electrons in12orbitalsandincludes:thelonepairandthe radicalcenteroftheOHradical,theσ andσ∗_and

theπ andπ∗orbitalsoftheCH3C=OOHdouble

bond,theσ andσ∗orbitalsof theCH3COO−H

bond,theσ andσ∗orbitalsofCH3CO−OH,and

thetwolonepairsof theaceticacidoxygens.As theactive spaceis ratherlarge, itwas not possi-bletodeterminethesaddlepointstructureatthis level of theory. Therefore M06-2X/aug-cc-pVTZ structures were usedin thecalculations. The en-ergy barriers,computed atthe CASPT2level on DFTgeometrieswithrespecttothereactantskept at10A˚ usinganIPEAshiftof0.25,are2.21and 0.71kcal/molforTSAandTSB,respectively.The calculatedtotalrateconstantforH-abstractionby OHfromCH3COOHiscomparedwith

experimen-talandtheoreticalestimatesinFig. 4 ,whilesaddle pointgeometriesandthereactionPESarereported inFig.S4.

Fig.4. Comparisonbetweenexperimental[14,15] ,points andcalculated[19,thiswork,1atm],lines,rateconstants forthetotalH-abstractionfromCH3COOH.Thewell skippinglabelmeansthatthecontributiontotherate con-stantfromthecollisionallystabilizedvanderWaalswell wasneglected.

Rateconstants werecomputedbothincluding and neglecting (well skipping) the contribution fromthecollisionallystabilizedvdWreactantwell summingTSAandTSBreactionfluxes.Itwasthus found thatthe inclusion of reacting fluxes from thevdW wellisof keyrelevancein orderto de-scribeproperlythelowtemperaturedependenceof therateconstantanditsupportstheclaimthatthe rateispressuredependentatlowtemperatures[16] . Therapidincreaseoftherateconstantabove500K isdeterminedbothbyreactingfluxesthroughTSA becoming fasterthanthose throughTSBandby theincreaseofrelevanceofH-abstractionfromthe methylgroup.Thedominantreactionchannelis H-abstractionof the acidic hydrogen upto 850K, aboveH-abstractionfrommethylbecomesslightly dominant.

RateConstantsinterpolatedinthemodified Ar-rheniusform,structures,energybarriers,hindered rotorPES,vibrationalfrequenciesofallstationary pointsarereportedinTableS6foreachinvestigated reactionchannel.

3.3. Kineticsimulations

ThePOLIMIkineticmodel,updatedwiththe acetic acid rate constants here determined, was usedtosimulatetheshocktubeexperimentaldata of Mackie et al. for the thermal decomposition of acetic acid in the 1300–1950 K temperature range.The experimentaldataarecomparedwith those calculated using the present mechanism, the Aramcomechanism [25,26] ,and the mecha-nismproposedbyChristensenandKonnov[31] in

Fig. 5 .

The agreementisgood forwhatconcernsthe prediction of acetic aciddecomposition andthe

formationofCO2,whileCH4compositionprofiles

are slightly underestimated, CH2CO is

overesti-mated,andCO,H2,aswellasC2 speciesprofiles

areunderestimated.Withrespecttotheliterature, thepresentmodelperformsgenerallybetter than theChristensenetal.modeland,forwhatconcerns CO, H2, andC2 species slightly worse than the

Aramco mechanism. The different performance of theAramcomodel forthementioned species ismostlyduetoitsinclusionof afastaceticacid decomposition channel to CO+OH+CH3. At

1600Ktherateof thischanneliscomparableto thatof reactionsR1 andR2.This isincontrast withourestimateoftherateof decompositionof aceticacidtoCH3+COOH,whichat1600Kis

afactorof3slowerthantherateusedinAramco andtherateoftheotherdecompositionchannels. Thediscrepancyismostlikelyduetothefactthat Aramcouses high pressure rate constants, while

our ME simulations show that at 1600 K and

1atmthehomolyticdecomposition channelisin significantfalloff.

Thereactionfluxanalysis,reportedinFig.S6,

shows that at 1600 K several decomposition

pathways areactive foraceticacid: 36% decom-posestoH2OandCH2CO,andasimilaramount

to CH4 and CO2, which are thus primary

de-compositionproducts.Theremaining decomposi-tion pathways are H-abstraction (20%) and

ho-molyticdecomposition toCH3 and COOH. CO

isformedthroughseveralchannels,mostly involv-ing the formation of CH2CO as an

intermedi-atespecies.The CH2COOHradical playsan

im-portant role in controlling the system reactivity in the present kinetic model, since recombining with H it can eitherterminate the radical chain formingCH3COOH,orpropagatedecomposingto

CH3 andCOOHorOHandCH2CO.The

dom-inant CH2COOH reaction channel is

decompo-sition to CH3 andCOOH, which requires

over-comingabarrierof41.9kcal/moltoisomerizeto thereactiveCH3COOintermediate.The

bimolec-ularreactivityofCH2COOHisatpresentmostly

unknown,madeexceptionfortheHaddition reac-tion,whichisthebackwardprocessofCH3COOH

decompositionandforwhichitwasthenpossible todeterminechannelspecifictemperatureand pres-suredependentrateconstantssolvingtheMEon theCH3COOHPES.The calculatedratesare

re-portedinTableS2.Additionalchannelsinvolving CH2COOHthatmayplayanimportantrolein

con-trollingthe systemreactivityare theaddition of CH3,OH,andOOH toCH2COOH. In

particu-lar,CH3 additionfollowedbyC2H5+COOH

de-composition may enhancethe rate of formation of C2 species anddecrease the CH2CO

concen-tration,thusprovidingapossibleexplanation of thereasonwhythesespeciesareoverestimatedand underestimated,respectively,inthepresentkinetic simulations.InadditionwenotethatCH2CO

py-Fig.5.Comparisonbetweenexperimentaldata(symbols[6] )andmodelsimulations(lines)fortheshocktubethermal decompositionof5%aceticacidinAr,at1atmand_τ–0.1ms.Redlinesrepresentfuelconversion,blacklinesrepresent intermediatesandproducts.

Fig.6. Comparisonbetweenexperimentaldata(symbols[33] )andmodelsimulations(lines)foraceticacidcombustionin laminarpremixedflames.

rolysis andcombustionarewellpredictedby the presentmodel (Fig.S7). Also,thesimulationsof the CH3COOH pyrolysis study of Sebbar et al. [32] showthatatlowertemperatures(1023K)there isgoodagreementbetweenmodelandexperiment forwhatconcerns aceticaciddecomposition and keteneformation(Fig.S8).

LaminarflamespeedspredictedbythePOLIMI model,comparedwithexperiments[32] inFig.S9, areontheaverageoverestimatedbyabout4cm/s. Thisisasignificantimprovementwith respectto theoriginalPOLIMImodel,whichoverestimated the experimental data by 9–10 cm/s, and with respect to the POLIMI model using the acetic acid decomposition kinetics implemented in the Aramcomodel(Fig.S10).Thesensitivityanalysis reportedintheSMshowsthisisindependent on aceticaciddecompositionkinetics.

Finally,thePOLIMIupdatedmodelwasusedto simulatethecombustionofaceticacid. Experimen-taldataandsimulationsarecomparedin Fig. 6 . Asfoundforaceticacidpyrolysis,therateof de-composition of acetic acid is well predicted by themodel,aswellasCO2,H2O,andCOprofiles.

Keteneisunderpredicted, whilemethaneis over-predicted.Thesedeviationsarenotconsistentwith whatfoundinpyrolysisconditions.Additional sim-ulationresultsarereportedinFig.S11,whilethe re-sultsofthesimulationof theexperimentalresults ofElwardanyetal.[9] arereportedinFig.S12.

4. Conclusions

Rateconstantsforthedecompositionof acetic acidthroughunimolecularandH-abstraction

reac-tionswereinvestigatedtheoretically.Thecalculated elementary reaction ratesare in goodagreement withexperimentaldata.Simulationof aceticacid pyrolysisandcombustionshowedthattheadopted kineticmodelisabletopredictwellaceticacid de-compositionandtheformationofitsmainprimary productsinalltheinvestigatedsystems.Theonly exception isketene,whose formationiswell pre-dictedatlowtemperatures,butunderpredictedin combustion andslightly overpredictedin pyroly-sis.Inaddition,theformationofsecondaryspecies, inparticularC2speciesandCOandH2in

pyroly-sis,isunderpredicted.Thesystemreactivity,in par-ticulartheflamespeed,issensitivetothekinetics oftheCH2COOHreactionintermediate,whichis

formedafterH-abstractionfromthemethylgroup ofaceticacid,andtoketenekinetics.Bimolecular reactionsinvolvingCH2COOHandotherradicals

thatarepresentinhighconcentration,suchasCH3

andOH,mayplayanimportantroleincontrolling thesystemreactivityandinfluencetheflamespeed, whichiscurrentlyoverpredicted.

Furtherexperimentalstudieswouldalsobe ben-eficial,sincethetwomostrecentliterature exper-imental CH3COOH pyrolysis studies cannot be

modeledusingthesamesetofdecompositionrates, whichmustbemodifiedbyafactorof3ormoreto fitthedata.

Acknowledgments

The authors acknowledgethe support bythe

European Union’s Horizon 2020 research and

innovation program (Residue2Heat, G.A. No.

654650 ).TheauthorsaregratefultoEliseoRanzi forcontinuousandfruitfuldiscussions.

Supplementarymaterials

SM-1: Rate constants, pressure dependence, kinetic simulations; SM-2:Reaction Mechanism; SM-3:Masterequationinputs.

Supplementarymaterialassociatedwiththis ar-ticlecanbefound,intheonlineversion,atdoi:10. 1016/j.proci.2018.06.137 .

References

[1] A. Oasmaa , Y. Solantausta , V. Arpiainen , E. Kuop- pala , K. Sipilä, Energy Fuels 24 (2010) 1380–1388 . [2] A. Fukutome , H. Kawamoto , S. Saka , J. Anal. Appl.

Pyrolysis 108 (2014) 98–108 .

[3] K. Onarheim , Y. Solantausta , J. Lehto , Energy Fuels 29 (2015) 205–217 .

[4] C.H. Bamford , M.J.S. Dewar , J. Chem. Soc. (1949) 2877–2882 .

[5] P.G. Blake , G.E. Jackson , J. Chem. Soc. B (1968) 1153–1155 .

[6] J.C. Mackie , K.R. Doolan , Int. J. Chem. Kinet. 16 (1984) 525–541 .

[7]K. Saito , T. Sasaki , I. Yoshinobu , A. Imamura , Chem. Phys. Lett. 170 (1990) 385–388 .

[8]N. Butkovskaya , G. Manke , D. Setser , J. Phys. Chem. 99 (1995) 11115–11121 .

[9]A. Elwardany , E.F. Nasir , Et. Es-sebbar , A. Farooq , Proc. Combust. Inst. 35 (2015) 429–436 .

[10]M.T. Nguyen , D. Sengupta , G. Raspoet , L.G. Vanquickenborne , J. Phys. Chem. 99 (1995) 11883–11888 .

[11]X. Duan , M. Page , J. Am. Chem. Soc. 117 (1995) 5114–5119 .

[12]J.M. Clark , M.R. Nimlos , D.J. Robichaud , J. Phys. Chem. A 118 (2014) 260–274 .

[13]W. Sun , T. Tao , R. Zhang , et al. , Combust. Flame 185 (2017) 173–187 .

[14]V.G. Khamaganov , V.X. Bui , S.A. Carl , J. Peeters , J. Phys. Chem. A 110 (2006) 12852–12859 .

[15]E. Szabo , J. Tarmoul , A. Tomas , C. Fittschen , S. Dobe , P. Coddeville , React. Kinet. Catal. Lett. 96 (2009) 299–309 .

[16]Y.W. Huang , T.J. Dransfield , J.G. Anderson , J. Phys. Chem. A 114 (2010) 11539–11544 .

[17]F. De Smedt , X.V. Bui , T.L. Nguyen , J. Peeters , L. Vereecken , J. Phys. Chem. A 109 (2005) 2401–2409 .

[18]D. Vimal , P.S. Stevens , J. Phys. Chem. A 110 (2006) 11509–11516 .

[19]J. Mendes , C.-W. Zhou , H.J. Curran , J. Phys. Chem. A 118 (2014) 12089–12104 .

[20]H.-J.Werner,P.J.Knowles,R.Lindh,F.R.Manby, M. Schutz, P. Celani, T. Korona, G. Rauhut, R.D. Amos, A. Bernhardsson, A. Berning, D.L. Cooper, M.J.O. Deegan, A.J. Dobbyn, F. Eckert, C. Hampel, G. Hetzer, A.W. Lloyd, S.J. McNicholas,W.Meyer,M.E. Mura,A. Nick-lass,P.Palmieri,U.Schumann,H.Stoll,A.J.Stone, R.Tarroni,T.Thosteinsson.http://www.molpro.net , 2010.

[21]S. Sharma , S. Raman , W.H. Green , J. Phys. Chem. A 114 (2010) 5689–5701 .

[22]Y. Georgievskii , J.A. Miller , M.P. Burke , S.J. Klip- penstein , J. Phys. Chem. A 117 (2013) 12146–12154 . [23]M.Frisch,G.Trucks,H.Schlegel,etal.,Gaussian

09,revisionA.2,(2009).

[24]C.Cavallotti,M.Pelucchi,Y.Georgievskii,S.J. Klip-penstein,(inpreparation)2018.

[25]W.K. Metcalfe , S.M. Burke , S.S. Ahmed , H.J. Cur- ran , Int. J. Chem. Kinet. 45 (2013) 638–675 . [26]S.M. Burke , U. Burke , R. Mc Donagh , et al. , Com-

bust. Flame 162 (2015) 296–314 .

[27]E. Ranzi , A. Frassoldati , A. Stagni , M. Pelucchi , A. Cuoci , T. Faravelli , Int. J. Chem. Kinet. 46 (2014) 512–542 .

[28]B. Ruscic , Int. J. Quantum Chem. 114 (2014) 1097–1101 .

[29]A. Cuoci , A. Frassoldati , T. Faravelli , E. Ranzi , Com- put. Phys. Commun. 192 (2015) 237–264 .

[30]J.M. Anglada , J. Am. Chem. Soc. 126 (2004) 9809–9820 .

[31]M. Christensen , A. Konnov , Combust. Flame 170 (2016) 12–29 .

[32]N. Sebbar , J. Appel , H. Bockhorn , Combust. Sci. Technol. 188 (2016) 745–758 .

[33]N. Leplat , J. Vandooren , Combust. Flame 159 (2012) 493–499 .