Evaluating the possibility of high-pressure desorption of CO2 via volatile co-solvent injection

ContentslistsavailableatScienceDirect

Chemical Engineering Research and Design

j o u r n a l ho me p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c h e r d

Evaluating the possibility of high-pressure

desorption of CO ₂ via volatile co-solvent injection

Ricardo R. Wanderley, Hanna K. Knuutila

DepartmentofChemicalEngineering,NorwegianUniversityofScienceandTechnology(NTNU),NO-7491 Trondheim,Norway

a r t i c l e i n f o

Articlehistory:

Received25November2020 Receivedinrevisedform4March 2021

Accepted11March2021 Availableonline17March2021

Keywords:

CO2capture CO2desorption CO2stripping High-pressure Water-leansolvent

a bs t r a c t

Thisworkevaluatesthepossibilityofemployingavolatileco-solventinjectionforrecover- ingCO₂fromloadedmonoethanolamineat120^◦Cunderpressuresabovethoseachievable throughregulardesorptionprocesses.Thisco-solventwouldbefeddirectlyintothereboiler, percolatingthecolumnanddeliveringhigheroperationalpressureswithoutsignificatively affectingthechemicalequilibriumbetweenCO2andtheamine.Removalofthisco-solvent wouldberequiredbeforetheleanamineisrecirculatedtotheabsorber.Ashortcutmethod- ologyforscreeningpossibleco-solventcandidatesispresented,andMESHcalculationsof hypotheticalstrippingprocessesemployingthehigh-pressuredesorptionapproachareper- formedtoillustratetheexpectedbehaviorofthesesystems.Pressuresabove500kPaare theoreticallyobtainablethroughtheuseofco-solventswhicharelessvolatilethanCO2but thatarestillgasesat25^◦Cand101.325kPa,suchasisobutaneanddimethylether.Theseco- solventswillleavethedesorberfractionedbetweenthedistillateandthebottomproduct, thusrequiringtwoadditionalseparationprocessforrecovery.Lessvolatilesolventswillcon- centrateatthebottomstagesofthedesorber,whilemorevolatilesolventswillflowstraight throughthecolumnallthewayuptothedistillatewithouteffectivelydeliveringpressuresas highasdesired.Inotherwords,thismethodologyresultsinadelicateoptimizationproblem offindingidealvolatilitiesandoperationalconditions.Thoughnodetailedenergyanalysis isperformedinthispreliminaryassessment,wehaveidentifiedapromisingopportunity forCO2productionathigherpressuresandenumeratedtheissuesoneshouldbeconcerned withwhenlookingfurtherintohigh-pressuredesorption.

©2021TheAuthors.PublishedbyElsevierB.V.onbehalfofInstitutionofChemical Engineers.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.

org/licenses/by/4.0/).

1. Introduction

Chemicalabsorption viaaqueous aminesolventsisanestablished technologyforCO2recoveryfromgaseousstreamssuchasnaturalgas, syngasandfluegas(Rochelle,2009).Ingeneralterms,thistechnique reliesonmanipulatingthetemperature-dependentchemicalequilib- riumbetweenamineandCO2forcapturingCO2atlowtemperatures inanabsorbercolumnandreleasingitathightemperaturesinastrip- percolumn.Whilemostcyclesworkwiththeabsorberdesignedto operateatabout30–40^◦Cregardlessofthesolvent(unlessitisapar- ticularlyvolatileone,suchastheAmisol^®solvent(Kriebel,1984),and ignoringthetemperatureincreasesbroughtbytheexothermicityof

∗ Correspondingauthor.

E-mailaddress:[email protected](H.K.Knuutila).

CO2absorption),thetemperatureofthestripperisneatlydelimitedby thethermaldegradationfeaturesoftheamine(Rochelle,2012,2016).

Intheparticularcaseofaqueousmonoethanolamine(MEA),themax- imumoperationaltemperatureofthereboilerisaround120^◦C(Vega etal.,2014).DavisandRochelle(2009)havedemonstratedthat,foreach 17^◦Cincreaseintemperature,MEAdegradationratesacceleratefour- fold.Afterbeingrecovered,CO2mustbecompressedtoupto6–10MPa fortransportationandinjection(Wangetal.,2019).Ithasbeenshown thathigherdesorberpressuresresultinlowercompressiondutiesto achievesuchtransportationconditionstogetherwithlowerregenera- tionduties(OyenekanandRochelle,2007).Andyet,reboilerpressure andtemperatureareintrinsicallyinterlinkedthroughthevapor-liquid equilibrium(VLE)behavioroftheaqueousaminesolvent.Ifonewants toobtainaleansolventwithadeterminedconcentrationofCO2below adeterminedtemperaturethreshold,thepressurecapisinherently fixed.

https://doi.org/10.1016/j.cherd.2021.03.011

0263-8762/©2021TheAuthors.PublishedbyElsevierB.V.onbehalfofInstitutionofChemicalEngineers.Thisisanopenaccessarticle undertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).

Thepresentstudyproposesanalternativetobreakthisdeadlock, whichistheinjectionofavolatileco-solventdirectlytothereboiler.The additionofanewcomponentintroducesanotherdegreeoffreedomto thestripperandallowsforCO2recoveryathigherpressuresunderthe sametemperaturethresholdof120^◦C.Thisco-solventwouldthenbe recoveredfromtheproductstreamsofthedesorber,withthechosen recoverytechnologybeingdependentonwhethertheco-solventleaves asagastogetherwithCO2orasaliquidmixedwiththeleanaminein thebottomproduct(oralternativelyfractionedbetweenbothstreams).

Inanycase,ourinitialconceptisthatthisco-solventshouldnotbe allowedtoreturntotheabsorbercolumnwiththeamine,whereit mightderangeabsorptioncapacitiesandratesasshowninourprevi- ousstudies(Wanderleyetal.,2019,2020)andadditionallycomplicate emissioncontrolsystems.

Thisisnotanentirelynewproposal.Asimilarconceptcanbefound intheworkofTobiesenandSvendsen(2006).However,thatstudywas fixatedontheideathatonemightbeabletodecreasereboilerduties withtheadditionofavolatileco-solvent.Asshownbytheauthors, thatisnotentirelyfeasible.Conversely,directstrippingoftheamine withaco-solventvaporstream(pentane)hasbeenanalyzedbyYang etal.(2020)andevaluatedintermsofregenerationduties,withno assessmentofthepossibilityofrecoveringCO2athigherpressures.

Ourapproachisslightlydifferentfromtheseworks:wedonotintendto decreaseregenerationdutieswiththisprocessmodification,andfully acceptthatthosewillprobablyincrease.Nevertheless,iftheaddition ofavolatileco-solventcanattainenoughpressure,thenthisincrease inregenerationdutiesmightbeoffsetbyadecreaseincompression duties.

Processdesignandsimulationarethetoolsuponwhichthiswork willrely toevaluate theconsequences ofemploying avolatileco- solventforhigh-pressuredesorption.Inthat,theapproachadopted hereisverysimilartothatemployedpreviouslyintheassessment ofhypotheticalwater-leansolventsforCO2 capture(Wanderleyand Knuutila,2020).Thewayinwhichthisstudyisstructuredisthefollow- ing:

i ThemethodologiestocalculateVLEbehaviorandsolvemassand energybalancesinadesorberoperatingatsteadystatearedescribed inSections2.1and2.3. Section2.2employstheVLE calculation toproposeashortcut methodologyforevaluatingcandidateco- solventswith theaid ofa simpledatabase containingAntoine parametersofseveralchemicalcompounds.

ii Shortcutevaluationsofpossiblevolatileco-solventsarecarriedout inSection3.1.Apatternisclearlyidentifiedinwhichmostplausible candidatesarelight,flammableorganiccompounds.

iii Desorberoperationswiththeadditionofaseriesofco-solvents thatareliquidat101.325kPaand25^◦Caresimulatedandana- lyzedinSection3.2.Theseareco-solventsofmoderatevolatility wellexemplifiedbytheseriesoffurans.

iv Desorberoperationswiththeadditionofaseriesofco-solventsthat aregasat101.325kPaand25^◦CaresimulatedandanalyzedinSec- tion3.3.Theseso-calledhypervolatileco-solventsarerepresentedby dimethyletherandisobutane.

Despitethetargetofthisprocessmodificationbeingthereduction oftotalpowerusageintheCO2captureplant,wemustremarkthatthis studydoesnotintendtocarryoutaproperenergeticorexergeticevalu- ationofhigh-pressuredesorptionasawhole.Therearepresentlymany unknownsregardingtheprocess,andthemethodologyemployedhere istoosimplistictocorrectlyestimatethecostsofco-solventseparation fromtheleanamineorevencorrectlyevaluateiftheliquidproductwill beasingle-phaseorabiphasicstream.However,thismethodologyis abletopredictandidentifypatternsandphenomenathatmightbe observedwhenemployingco-solventinjectionforhigh-pressuredes- orption.Therefore,thisisavaluablepreliminarystudyinapossible futureCO2recoverytechnology.

2. Methodology

Forthefollowingseriesofdevelopments,wehaveemployed anequilibriumapproachthatconcealsthereactionsbetween CO₂,MEAandwater.Thisimpliesthatmassbalancesthrough- outthecolumnwillkeeptrackofMEAconcentrationswhile concealingthefactthatthe‘MEA’subscriptactuallystandsfor amixtureoffreeMEA,protonatedMEAandMEAcarbamate.

Similarly,massbalanceswillkeeptrackof‘CO2’ concentra- tionsintheliquidphasewhileconcealingthatthesevalues applyformolecularCO₂plusMEAcarbamateandbicarbon- atemolecules.Onecouldpointoutthat,e.g.,thismeanswe arecountingMEAcarbamatetwice.Inreality,whatmatters is that, through consistentchecking ofmass balances and equilibriumcalculations,allmasstransferphenomenainthe desorberarethoroughlyaccountedfor.

Thevapor-liquidequilibriumofCO₂inthesolventisgiven bythesoftmodel,whichcorrelatestheCO2partialpressurein thevaporphaseinkPa(pCO2)withCO2loadinginmolCO2/mol amine (˛)andtemperature (T)inK (Aronu etal., 2014).For aqueousMEA30%wt.,thesoftmodelequationisdefinedas Eqs.(1a)–(1d).

ln(pCO2)=1.8·ln(˛)+k1+ 10

1+k2·exp(k3·ln(˛)) (1a)

k1=−9155.955·1

T+28.027 (1b)

k2=exp

−6146.18·1 T+15

(1c)

k3=−7527.0376·1

T+16.942 (1d)

Meanwhile,thevapor-liquidequilibriaofwater,MEAand theco-solventarecalculatedbyRaoult’slawinconjunction withDalton’slaw,Eq.(2).InEq.(2),pisthetotalpressure,pisat

isthesaturationpressureofcomponenti,andy_i andx_iare respectivelythemolarfractionsofcomponentiinthevapor andintheliquidphase.

p·y_i=p^sat_i ·x_i (2)

Theconcentrationsofamineandwaterbeforetheaddition oftheco-solventarespecifiedbythefactthatweareoperat- ingwithaqueousMEA30%wt.,sothatinitiallyxMEA=0.1122 andx_H2O=0.8878.Theadditionofaco-solventsimplyimplies renormalizingthese molarfractions. Wehaveintroduceda factorfCOSwhichaccountsforhowmuchco-solventissolubi- lizedintheliquidphaseasaratioofthewatercontentofthe freshsolvent,Eq.(3).

f_COS= xCOS

x^fresh_H2O

(3)

Additionally,theabsorptionofCO2bringsafourthcompo- nenttothemixture,andtheloading˛canbeemployedasa secondrenormalizationfactor.Renormalizingthemolarfrac- tionsofwater,MEA,co-solventandCO₂entailsEqs.(4a)–(4d).

x_H2O= x^fresh_H2O

(1+f_COS)·x^fresh_H2O +(1+˛)·x^fresh_MEA

(4a)

xMEA= x^fresh_MEA

(1+f_COS)·x^fresh_H2O +(1+˛)·x^fresh_MEA (4b)

xCOS= f_COS·x^fresh_H2O

(1+fCOS)·x^fresh_H2O +(1+˛)·x^fresh_MEA

(4c)

x_CO2= ˛·x_MEA^fresh

(1+fCOS)·x^fresh_H2O +(1+˛)·x^fresh_MEA

(4d)

Theprevious Eqs.(4a) to (4d) are useful for calculating liquidphasemolarfractionconcentrationsformassbalance purposes.However,theymightbeinadequateforvaporpres- surepurposes,i.e.forusingEq.(2),particularly inthe case ofwaterandMEA.Thereasonisthat, asCO2 isconsumed bytheliquidphaseandthe loadingincreases,theeffective molarfractionsofamine andwaterwillbereduced.If one weretoconsiderthereactionbetweentwomoleculesofamine withoneofCO₂togenerateonemoleculeofcarbamateand protonatedamine,therelationshipbetweenCO₂loadingand thenumberofmolsofaminewouldbeprettystraightforward, Eq.(5a).Additionally,theconcentrationofwaterintheliquid phasewouldbeindependentofloading,Eq.(5b).

n^eff_MEA=n^app_MEA·(1−2·˛) (5a)

n^eff_H2O=n^app_H2O (5b)

TheproblemwithEq.(3)isthatitdoesnotapplyforhigher CO2loadings,wheretheparticipationofwaterinthereaction mechanism becomesmorerelevant through the formation ofbicarbonateandcarbonate.Inthatcase,Eqs.(4a)and(4b) becomeinadequate.

Wonget al. (2016) have publishedRaman spectroscopic dataforthespeciationofthereactivewater–MEA–CO2milieu at40^◦C.Thisdatawasemployedtofitthedegreesofadvance- ment(,whereisavectorwithcomponents1,2and3)of thefollowingsetofreactionsasafunctionof˛:

H₂O_(l)+MEA_(l)+CO_2(g)→₁HCO⁻_3(l)+MEAH⁺_(l) (R1)

H2O_(l)+2·MEA_(l)+CO_2(g)→2CO⁻_3(l)²+2·MEAH⁺_(l) (R2)

2·MEA_(l)+CO_2(g)→3MEACOO⁻_(l)+MEAH⁺_(l) (R3)

ThissetofequationsignoresthepresenceofmolecularCO2

intheliquidphase,whichisnonethelessquitesmallunder loadingsof˛= 0.6(Wong etal., 2016),especiallyathigher temperatures.Oncethevaluesofarefound,thenumberof freewaterandMEAmoleculescanbecalculatedasafunc- tionof˛.Noticethat,throughreactions(R1)–(R3),thenumber ofmoleculesintheliquidphaseisnevermodified.Therefore, Eqs.(5a)and(5b)shouldbevalidbothforthecalculationof molenumbersasforthecalculationofmolarfractions,asno renormalizationisnecessary.Inthecaseoftheadditionofa co-solvent,theapparentconcentrationofwaterandMEAcan beconvenientlycalculatedbyEqs.(6a)and(6b).

x^app_H2O= x^fresh_H2O

(1+f_COS)·x^fresh_H2O +x^fresh_MEA

(6a)

Fig.1–MolarfractionsofwaterandMEAasafunctionof loading,derivedfromdataforaqueousMEA30%wt.at40

◦CobtainedbyWongetal.(2016).

x^app_MEA= x^fresh_MEA

(1+fCOS)·x^fresh_H2O +x^fresh_MEA

(6b)

Thefittingofhasbeenperformedbytheparticleswarm optimization described in past works (Evjen et al., 2019;

Skylogiannietal.,2019),andwiththisfittingwewereableto obtainanexpressioncorrelatingtheeffectivex_H2Oandx_MEAto

˛,Eqs.(7a)and(7b).Avisualinterpretationoftheseequations isshowninFig.1,wherefMEA(˛)istheexpressionshownin Eq.(7a)andfH2O(˛)istheexpressionshowninEq.(7b).

x^eff_MEA=x^app_MEA·exp

−0.0451−1.9910·˛−3.4911·˛² +3.6741·˛³

(7a)

x^eff_H2O=x^app_H2O·exp

−0.0069+0.0971·˛−0.2983·˛² +0.1665·˛³

(7b)

Fig.1showsthat,forloadingsbelow˛=0.4,theapprox- imationthattwomoleculesofMEAare consumedforeach moleculeofCO2absorbedisactuallyfine.Thiscanbeseenin howtheboldredandbluelinesapproachthedashedblack lines, which account solelyfor the carbamate mechanism.

Thismechanismaloneisnotenoughtoexplainhowwater andamineareconsumedathigherloadings,andEqs.(7a)and (7b)becomemorerelevant.

WithEqs.(6a),(6b),(7a)and(7b),Eq.(2)canberewritten asEqs.(8a)and(8b).Eq.(8c)showshowtocalculatethepar- tialpressure oftheco-solventinthevapor phase.Inthese expressions,weare alsomaking itclearthatthefreshsol- vent,bydefinition,consistssolelyofwaterandMEA,i.e.xH2O

=1–xMEA.Theconsequenceisthat,ifonesettlesforaninitial concentrationofaqueousunreactedMEA,aCO2loading,aco- solventfactorfCOSandatemperature,theequilibriumsystem iscompletelydefined.Asmentionedbefore,theCO₂partial pressurecomesfromEq.(1a),whichdependsonlyonloading andtemperature.

pH2O=p^sat_H2O·x^eff_H2O

=p^sat_H2O· 1−x^fresh_MEA (1+fCOS)·

1−x^fresh_MEA

·fH2O(˛) (8a)

pMEA=p^sat_MEA·x^eff_MEA

=p^sat_MEA· x^fresh_MEA (1+f_COS)·

1−x^fresh_MEA

+x^fresh_MEA

·fMEA(˛) (8b)

pCOS=p^sat_COS·x^eff_COS=p^sat_COS·

f_COS·

1−x^fresh_MEA

(1+fCOS)·

1−x^fresh_MEA

As a sidenote: Eqs. (7a) and (7b) are useful as a step- pingstoneforcalculatingvaporpressuresviaEqs.(8a)–(8c).

However,wemust highlight the factthat theyintroduce a

‘destruction’ofamineandwaterspecieswithincreasedCO2

loadingsthatwouldberatherproblematicforkeepingmass balancesthroughoutthecolumnincasetheywereusedfor evaluatingliquidphaseconcentrationsinsteadofEqs.(4a)and (4b).Inotherwords,thesetofEqs.(4a)–(4d)hasbeendesigned sothatthesumofallmolarfractionsintheliquidphaseis alwaysx_i =1,whereas Eqs.(7a)and (7b)donotobeythis rule.Therefore,onemustbecarefultodistinguishwhereEqs.

(7a)and(7b)areapplicableandwheretheyarenot.

Thesaturationpressureofwater,amineandco-solventcan becalculatedthroughtheAntoineequation,Eq.(9).Eq.(9)is writteninaformwhereinthetemperatureisgiveninKand thesaturationpressureisdeliveredinkPa,thoughitsparam- etersA_i,B_i andC_i havebeenobtainedinadatabasewhich requiresTin^◦Candp^sat inmmHg.TheAntoineparameters forwater,MEAandsomecandidateco-solventsaregivenin theAppendixAofthisstudy.

log₁₀

=A_i− B_i

C_i+T−273.15 (9)

Thoughthis formulationofthe vapor-liquidequilibrium problemmightseemconvoluted,itisactuallyveryconvenient.

WhileEqs.(4a)–(4d)offerasimplewayofkeepingtrackofthe flowratesofallcomponentsintheliquidphase,Eqs.(1a)and (8a)–(8c)offerawayofkeepingtrackofflowratesinthevapor phase.Theapplicationoftheseformulaewillbeshowninthe followingSections2.2and2.3.However,tofinishthissection, itmightbeinterestingtolisttheassumptionstakenduring thederivationoftheseequations.

i BothRaoult’sandDalton’slawsarevalidforfreeunreacted molecules,whichisreasonableduetotherelativelylow pressuresandhightemperatures.Thismeansthatfugac- itycoefficientsandactivitycoefficientsarealwaysunity, regardlessofloading;

ii The dependency between CO2 loading and CO2 partial pressure does not change with the addition of the co- solvent.Thiscanbearguedtonotbetrue,seeforexample Wanderleyetal.(2020);

iii Theadditionoftheco-solventdoesnotbringneitheranew reactionwithCO2nor anewreactionwithMEA,i.e.the co-solventmustbeperfectlyinert;

iv Additionally, the dependency between CO₂ loading and CO2partialpressurefollowsthemodelofAronuetal.(2014) parametrizedforaqueous30%wt.MEAregardlessofthe factthattheproportionsofwaterandamineareallowed tovaryinourcalculations;

v Theco-solvent isdeemed to besoluble inthe aqueous phase.Nosecond liquid phase formationisconsidered.

Acompendium ofwatermiscibility ofmany of the co- solventsexploredinthisworkcanbefoundinYaws(2003), thoughamine speciation willhave animpact inliquid- liquidequilibriaasseeninthecaseofbiphasicwater-lean solvents(Zhangetal.,2012,2019;Zhuangetal.,2016);

vi ThespeciationdataobtainedbyWongetal.(2016)foraque- ous30%wt.MEAat40^◦Cisvalidforvaryingwater-amine concentrationsevenathighdesorbertemperatures.

Oftheseassumptions,webelievethat(i),(iv)and(vi)are relatively inconsequential.Assumptions(ii),(iii)and (v)are slightly moreproblematic,and theyare discussed againin Section3.1.

2.2. Shortcutevaluationofco-solventcandidates

WiththeequationsshownintheSection2.1,theevaluation ofco-solventcandidatesisverystraightforward.Ifonefixes the concentrationofaqueousunreacted MEA(MEA30%wt.

impliesxMEA=0.1122),thedesiredleanloadingofthesolvent andthereboilertemperature,eachfCOSwillresultinadiffer- enttotalpressurep.ThisisshowninEq.(10),whichrelieson theformulaepresentedinSection2.1.

p=pH2O

x^fresh_MEA,˛,fCOS,T

x^fresh_MEA,˛,fCOS,T

(˛,T)+pCOS

x^fresh_MEA,˛,fCOS,T

(10)

In Section 3.1, we perform a screening of possible co- solventcandidatesbyfixingadesiredleanloadingof˛=0.2 molCO2/molMEA,areboilertemperatureof120^◦CandafCOS

=0.1molco-solvent/molwater.ForSections3.2and3.3,fCOS

isallowedtovarywhiletheremainderprocessspecifications arekeptjustasinSection3.1.Theimportantaspectofthis analysis isthatthe onlyparameters directlydepending on thenatureoftheco-solventarethethreeAntoinecoefficients usedtocalculatetheco-solventsaturationpressure,Eq.(9).By compilingacomprehensivedatabaseofAntoinecoefficients, oneisabletocarryout thisshortcut evaluationforalarge arrayofco-solventcandidates.Anexampleofthisprocedure isshowninSection3.1.

However, onemust notice that this shortcut methodol- ogydoesnotdirectlyindicatehowmuchco-solventmustbe injectedintheprocess.Theoperationalconditionsofthedes- orber(for example,its refluxandboil-upratios,RD and RB) affecttheliquidandvaporflowratesenteringandleavingthe reboiler.Theseflowratescannotbeobtainedwithoutafull assessmentofthestrippercolumn.Asaresult,thereisno straightforwardcorrelationbetweentherequiredmolarfrac- tionofco-solventinthereboilerliquidphase(x_COS=f_COS×x_H2O) andthemolarflowrateofco-solventthatmustbeinjectedto thecolumn(FCOS).Forco-solventsthatarenothypervolatile,

Fig.2–Schematicrepresentationofthedesorberwith co-solventinjection.

itisasuitableeducatedguesstoestimatethatallco-solvent injected ends up in the reboiler liquid phase, i.e. FCOS ≈ fCOS×FH2O,FEED. For hyper volatile co-solvents, the complete desorbermodellingisevenmoreessentialforunderstanding theperformanceofthestripperemployingco-solventinjec- tion.

2.3. DesorbermodellingwithMESHalgorithm

TheacronymMESH standsforMaterial, Equilibrium, Sum- mationandHeat (Wagialla andSoliman,1993).Modelling a desorberwithaMESHalgorithmimpliessolvingallmaterial andenergybalanceswhileapplyingvapor-liquidequilibrium equations to each stage. In our previous work, we have employedaMESHproceduretomodelanabsorberoperating withwater-leansolvents(WanderleyandKnuutila,2020).The procedureadoptedinthepresentstudyisquitesimilartothat one.Agooddescriptionofhowtoquicklyimplementandsolve MESHequationsisgivenbySteffenandSilva(2017).

Aschematicdrawingoftheproposeddesorptioncolumnis showninFig.2.

The desorber is modelled as having N + 2 equilibrium stages,meaningithasonecondenser,onereboilerandNinner stages.Thecondenserisapartialcondenserwhereanamount ofenergyQCisremovedfromthestageandavaporstreamV0

=Disobtainedasdistillate.Thereboilerisapartialreboiler whereanamountofenergyQRisaddedtothestageandaliq- uidstreamL_N+1=Bisobtainedasbottomproduct.Therich amine,whichisaqueousMEA30%wt.withaloadingof˛= 0.5molCO2/molMEA,isfedtothefirststageofthecolumn at105^◦CandamolarflowrateFFEED=1kmol/h.Astreamof pureco-solventisfeddirectlyintothereboilerat120^◦Cand amolarflowrateF_COS.EssentiallytheabsolutevaluesofFFEED

andF_COSarelessinterestingforsimulationpurposesthantheir valuesrelativetoeachother.Thecolumnismodelledasbeing perfectlyisobaric.

Notice that vapor-liquid equilibrium demands that the pressureofthereboilerbefixedonceitscompositionandtem- peratureare defined.Ourprocess requirementsare already specifying a maximum reboilertemperature and a desired solventleanloading.Therefore, ifonewantstomodifythe operational pressures ofthe process, the most straightfor- wardwayofdoingsoisbyshiftingtheconcentrationsdirectly inthereboiler.Thisiswhyweproposetheco-solventaddi- tionpreciselyintothatstage.Moreover,sincethisco-solvent isvolatile,aninjectionanywhereelseinthecolumnwould requireanincreasedlevelofco-solventcondensationinorder toaffectthevapor-liquidequilibriuminthereboiler,which wouldimplyincreasinglyhighcondenserduties,recirculation rates,andreboilerduties.

Solvingthismodelimpliessolvingthemassbalancesfor eachoneofthefourcomponentsacrosstheN+2stagesand thensolvingtheenergybalancesateachstage.Thismeans solving5×(N+2)equations.Additionally,tobeabletoadjust the operationalspecificationsofthedesorber,wearefixing onedesiredvariableatthetopandoneatthebottom.Atthe bottomofthecolumn,wehavespecifiedthattheleanload- ingofthesolventcomingoutmustbe˛=0.2molCO2/mol MEA.Atthetop,wehadtomakeadecisionbetweenspecify- ingthetemperatureofthecondenserortheCO2concentration inthe distillate. Wehaveultimately decidedtospecify the temperatureofthecondenser.

Asmentionedpreviously,thecolumnismodelledashav- ing the same pressure across all stages. Thispressure has beeninitiallysetasthatgivenbyEq.(10),i.e.asthatevalu- atedbytheshortcutmethod,beingessentiallydependenton theco-solventflowrateF_COS (orratheroftheestimatedf_COS atthereboiler).Due totheshortcomings ofthis roughcal- culation,it happenedoftenthattheMESHalgorithm,when fully solved,returnedatemperature atthe reboilerslightly superiororinferiorto120^◦C.Wehaveaddedasmallitera- tionlooptore-estimatethecolumnpressurepwiththeintent ofmeetingthereboilertemperatureof120^◦Cwithinamar- ginof±0.05^◦C.Therefore,thereboilertemperatureisalsoa specificationoftheprocess,althoughitisbeingmanipulated indirectlythroughtheupdatingofp.

Theenergybalancesinthisstudyhavebeenperformedin thesamewayasthoseofourpreviouswork(Wanderleyand Knuutila,2020).Theheatofvaporizationofwater,amineand co-solventhasbeenrecoveredfromtheAntoineexpression Eq. (9)through the use ofthe Clausius-Clapeyronrelation- ship. Theheatofvaporization ofCO₂ hasbeen considered constantat−H^abs=85kJ/molCO₂followingtheexperimen- tal dataobtainedbyKimet al.(2014) andWanderleyet al.

(2020)amongothers.Thegasheatcapacityhasbeencalcu- lated usingtemperature-dependentparametersprovidedby Yaws(2003)and,inthecaseofnitrogen,byCoker(2007).We havetakenthedecisiontoemployonlygasheatcapacitiesin thiswork.Thiscanbedonesince,followingthermodynamic consistency,theheatdemandedtochangethetemperatureof aliquidbetweenT1andT2istheheatdemandedtovaporize theliquidatT1,plustheheattoshiftthetemperatureofthe gasfromT1toT₂,plustheheatdemandedtocondensethegas atT₂.Assuch,thisapproachallowsustoavoidgettingbogged onhavingtomakeassumptionsregardinghowtocalculatethe heatcapacityofelectrolyticsolutions.Anobviouscriticismis thatthisisahighlysimplifiedwayofperformingcalculations.

Nevertheless,thismethodologyensuresthatthermodynamic consistencyisachievedintheenergybalances.Moreover,the goalofthisstudyisnottocomeupwithenergyvaluesforthe

performanceofhigh-pressuredesorptionsystems,butsimply toidentifypatternsandbehaviorswhenemployingvolatileco- solvents.Forthesakeofobtainingthesepatternsinatimely andcomprehensivemanner,webelievethatourapproachis goodenoughtosolvetheMESHequations.

Tofacilitatethediscussion ofour results,wewillintro- ducesomeusefulparameters.ThesearetherefluxratioRD, Eq.(11a),andtheboil-upratioRB,Eq.(11b).Additionally,we willdiscussthecondenserandreboilerdutiesQ_CandQ_R in termsofthe amountofCO2 recoveredinthe distillate, i.e.

thesedutieswillbegiveninMJ/kgCO2recoveredinsteadof, forexample,MJ/h.

RD=L0

D (11a)

RB= VN+1

B (11b)

Thestripperhasbeen proposedashaving5inner equi- libriumstages,meaning5+2intotal. Inouranalyses,the temperatureofthe reboiler of120^◦C and the lean solvent loading˛=0.2molCO2/molMEAwereselected.Thisleanload- ingwasconsideredtobesimilartovaluescommonlyfound inexperimentalandmodellingpapersregardingplantoper- ations withaqueousMEA, e.g.,Kvamsdal etal. (2009). The choiceof5innerequilibriumstageswassomewhatarbitrary, but isjustifiedbythe reboilerduties obtainedthroughour modellingapproach(seebelow).Anincreaseinthenumber ofequilibriumstageswouldmerelyslightlyreducetheenergy dutiesinthedesorberwithoutsignificativelyalteringanyof thetrendsobservedthroughoutoursimulations.

Withtheselectedleanloadingandreboilertemperature, themodelhasreturnedatotalpressureofp=188kPa.Fig.3 showssomeresultsforthedesorbermodellingwithoutthe additionofco-solvent.Thevaporphaseconcentrationprofiles forthespecificcaseinwhichthetemperatureofthecondenser isfixedatT_C=35^◦Careshownintheupper-leftcorner.With theseconditions,CO2isproducedatapurityofabout97%.In theupper-rightcornerofFig.3,onecanseehowincreasingthe temperatureofthecondenserallowsmorewatertobedrawn outasaproduct,dilutingtheCO₂stream.Highercondenser temperaturesimply lower reflux ratios(bottom-left corner) and lower condenser duties (bottom-right corner). Boil-up ratios and reboilerduties are alsoreduced with increasing condensertemperatures,butveryslightly.

Thereboiler duties evaluated forthis process (QR ≈ 3.5 MJ/kgCO₂)areverysimilartothoseobtainedinrealindus- trialCO2 captureapplications beforeprocess modifications suchasvapor recompressionand advancedflashstripping (Rochelle,2016).Thissuggeststhatourmodellingapproach, thoughhighlysimplified,issophisticatedenoughtodeliver credible dataonthe designand operationofCO₂ desorber columns.

3. Results and discussion

Forevaluatingaseriesofco-solventcandidates,oneneedsa gooddatabaseofAntoineparameters. Foraninitialassess- ment, the database compiled by Yaws and Satyro (2015a) showedtobeavery good resource,withmorethan 25,000 organic compounds. However, for the second step of this work,i.e.thedesorbersimulations,thisdatabaseprovedtobe

Table1–Organiccompoundsandthetotalpressure attainedinthereboilerforMEA30%wt.when˛=0.2, f_COS=0.1andT=120^◦C.CaseA:co-solventcanbe recoveredasliquidat25^◦Cand101.325kPa.

Name CAS p/kPa

2,3-butadien-1-ol 18913-31-0 959.6

Vinylformate 692-45-5 753.9

Hydrogencyanide 74-90-8 396.8

3-methoxy-1-propene 627-40-7 394.8

1,trans-2-dimethylcyclopropane 2402-06-4 374.3

Dimethylacetylene 503-17-3 369.3

Divinylether 109-93-3 357.9

3-methyl-1-butyne 598-23-2 350.5

(S)-(−)-propyleneoxide 16088-62-3 347.7 1,cis-2-dimethylcyclopropane 930-18-7 343.7

Table2–Organiccompoundsandthetotalpressure attainedinthereboilerforMEA30%wt.when˛=0.2, f_COS=0.1andT=120^◦C.CaseB:co-solventcanbe recoveredasliquidat25^◦Cand1013.25kPa.

Name CAS p/kPa

Propane 74-98-6 1082.1

Vinylalcohol 557-75-5 1057.2

Cyclopropane 75-19-4 1013.8

Cyanogen 460-19-5 969.3

Methylacetylene 74-99-7 962.2

2,3-butadien-1-ol 18913-31-0 959.6

Allene 463-49-0 866.2

Dimethylether 115-10-6 847.3

Methylamine 74-89-5 764.7

Vinylformate 692-45-5 753.9

unsuitable.Thisisbecauseoneneedsheatcapacitydatatocal- culatetheenergybalancesinsideadesorbercolumn,andheat capacitydataformanyofthecompoundscompiledbyYaws andSatyro(2015a)isquitedifficulttobefound.Therefore,for thesecondpartofthisstudy,thedatacompiledbyYaws(2003) hasbeenusedinstead.Thoughdisplayingasmallerdataset, comingslightlyshortof5000organiccompounds,thissecond resourcecompilesbothAntoineparametersandheatcapacity parametersforroughlythesamearrayofchemicals.

Atanyrate,employingthedatabaseofYawsandSatyro (2015a), onecan employtheshortcut calculationdescribed previouslyinSection2.2toestimatethetotalpressureofthe reboiler.Forthepresentcalculations,weareassumingthatthe leanamineshouldberecoveredat˛=0.2molCO2/molMEA and120^◦C.WehavealsofixedfCOS=0.1molco-solvent/mol waterforthesakeofsimplicity,thoughoneshouldnoticethat thereboilerpressurewillasymptoticallyapproachthesatura- tion pressureofthepureco-solventasf_COS becomeslarger.

Finally,wehaveseparatedouranalysisintwocases.InCase A,the co-solventcanberecovered asaliquidat25^◦C and 101.325kPa,meaningtheco-solventcanbeordinarilyrecov- eredbycooling.InCaseB,theco-solventcanberecoveredas aliquidat25^◦Cand1013.25kPa,meaningtheco-solventcan berecoveredaftercoolingandcompression.

Thetotalpressuresobtainedbyfollowingthesetwoanaly- seshavebeenrankedindescendingorderandtheresultsare showninTable1andTable2.Wetookthelibertyofmanually removingfromthelistallcompoundsthatpresentedhalogens (F, Cl, Br, I), phosphorus, selenium, silicon and any het- eroatomsotherthanoxygenandnitrogen.Thishasbeendone topreventclearcasesinwhichtheco-solventwouldpoten- tially acceleratethedegradationoftheamine(Moseret al.,

Fig.3–Desorbermodellingwithoutco-solvent.Vaporphaseconcentrationprofilesforcaseinwhichthecondenser temperatureis35^◦C,thendistillateconcentrations,refluxandreboilerratiosandcondenserandreboilerdutiesas functionsofcondensertemperature.

2011, 2019).That bringsameaningfultrimmingofoptions, as the vast majority of ranked co-solvents are chemicals suchasmethyl1,1,2,2-tetrafluoroethyl ether,fluorocarbonyl isocyanateandboranedimethylsulfide.Byeliminatingthese candidates,thelistisvastlyreduced.

Otherclearlyharmfulchemicalshavebeendeliberatelyleft inTable1andTable2forillustrationpurposes.Forexample, noonewouldrecommendusinghydrogencyanide(usedas chemicalweaponintheFirstWorldWar)orcyanogenasco- solventsduetotheirhightoxicity.Similarly,thepresenceof methylamineinTable2shouldraisetheconcernofwhether thisco-solventwouldreallyactasaninertgas–quiteproba- bly,itwouldreactwithCO2(Hajmaleketal.,2013).Therefore, the use of the shortcut method forevaluating co-solvents requiresjudgement.And ifthis appliestoobvious harmful chemicalssuchashydrogencyanide,italsoappliestoevery otherchemicalwhichmightbeunfamiliartothereader.The compilationandunderstandingofmaterialsafetydatasheets (MSDS)ofeachcomponentisessentialforevaluatingpossible co-solvents.

TheAppendixAofthis workincludesanexpandedver- sion of Table 1 and Table 2 including chemical structures andhazardsymbolsforeveryco-solventcandidate.Whatcan be summarized from a cursory study of the MSDS isthat alloftheprospectiveorganicco-solventsfoundthroughthis

methodologyareflammable.Thisisperhapsnotsurprising.

Onemustnotice,however,thatsomeofthemareextremely unstable.Forexample,neithercyclopropanenorvinylalco- holcanbeobtainedcommerciallyduetotheirhighinstability, withtheformerquicklycombustinginthepresenceofoxygen andthelatterbeingspontaneouslyconvertedtoacetaldehyde withinashortperiodoftimeafterproduction.Ethyleneoxide (whichshoulddeliver456.3kPaoftotalpressurefollowingour methodology)isso unstablethatonehandbookstatesthat

“Althoughsoluble inwater, solutions willcontinue toburn until diluted to approximately 22 volumes of water to one volume of ethyleneoxide”(Pohanish,2012).Thisbegsforcautionanddis- cernmentwhenconsideringtoapplyanyoftheco-solvents proposed inthis study inreal lifeapplications,whether in industrialorlaboratoryscalelevels.

Together with the MSDS, oneshould carefully consider thechancesoftheco-solventreactingwithCO₂orwiththe amine, since bothpossibilities would haveharmful effects on the desorption process. In ourprevious work,we have brieflydiscussedthereactivityofMEAwithdifferentorganic solvents(Wanderleyetal.,2020).Thesegeneralrulesdonot preclude the necessityof empiricaldataand experimental investigation.Everythingobtainedthroughshortcutmethods andmathematicalmodellingneedstobevalidatedbyscien- tificobservation.

Whatwouldhappeniftheco-solventisnotentirelymis- ciblewiththeaqueousamine?Inthatcase,asecondliquid phasewouldbeformedcontainingaproportionallyhighcon- centrationofthe co-solvent. Thissecond liquid phasestill hastobeinequilibriumwiththevaporphase,andthepar- tial pressure of the co-solvent will end up being actually largerthanitwouldappeartobethroughtheapplicationof asingle-liquidphase calculation. Additionally, animmisci- bleco-solventwouldprobablyfacilitatetherecoveryofthis compoundfromtheliquidleanamineproduct,moreeasily closingthe co-solventlooparound thedesorber.Therefore, theimmiscibilityoftheco-solventmightactuallybebenefi- cialtothe process.Conversely,if theco-solventismiscible anddoesaffectthechemicalequilibriumbetweenCO₂andthe amine,thecurrentunderstandingofVLEbehaviorinwater- leansolventsimpliesthatthisco-solventwillhelpdesorbthe CO2byshiftingthereactiontowardslesscarbamateformation (Wanderleyetal.,2019,2020;YuanandRochelle,2018,2019).

Thenet effect on the CO2 stripping process would, there- fore,alsobepositive.Atanyrate,theconcernsraisedatthe endofSection2.1showthatourmethodologyisactuallypes- simistictowardstheuseofvolatileco-solvents,andthusthat theresultsobtainedwithoursimulationsaregenerallyquite conservative.

Fig.4–Seriesoffuransandtheirsaturationpressuresat 120^◦C,calculatedwiththeAntoineparametersfromYaws (2003).

3.2. Desorberwithseriesoffuranderivatives

To exemplifythe effects ofadding a volatile co-solvent to the desorber, we introduce the following series of furans shown in Fig. 4. This series is presented in a descend- ing order of volatility, with furan being the most volatile co-solvent and 3-methyltetrahydrofuran being the least.

The saturation pressures of these components at 120 ^◦C have been calculated with the Antoine parameters pro- vided inYaws (2003) andare printed onFig. 4. Conversely, their boiling points at 101.325 kPa are calculated by the same approach as being 31.3 ^◦C for furan, 64.8 ^◦C for tetrahydrofuran, 80.2 ^◦C for 2-methyltetrahydrofuran and 138.0 ^◦C for 3-methyltetrahydrofuran. In other words, 3- methyltetrahydrofuranseemstobelessvolatilethanwater according to the parameters provided by Yaws (2003). We

Fig.5–Vaporphasemolarfractionsofwater,amine,CO2andco-solventinadesorberwithfuranandtetrahydrofuran injection.Desorberwith5+2equilibriumstages.Therichsolventmolarflowrateis1kmol/h.Richloading˛=0.5,lean loading˛=0.2,temperatureatthereboilerof120^◦Candtemperatureatthecondenserof35^◦C.Theco-solventmolarflow ratesareprintedonthegraphs.

Fig.6–Temperatureprofileswithfuranasco-solvent.

Desorberwith5+2equilibriumstages.Therichsolvent molarflowrateis1kmol/h.Richloading˛=0.5,lean loading˛=0.2,temperatureatthereboilerof120^◦Cand temperatureatthecondenserof35^◦C.

musthighlightthatthelaterversionofthisdatabookprovides drasticallydifferentparametersfor3-methyltetrahydrofuran (YawsandSatyro,2015a),andthatthissolventmightaswell bemorevolatilethanwater.Nevertheless,theseriesasitis shownisdeemedtobeillustrativeoftheeffectsofdecreasing volatilitiesofpossibleco-solvents.

Thevapor phase molarfraction profiles ofwater, MEA, CO2andco-solventareshownfortheadditionoffuranand tetrahydrofuranatdifferentflowratesinFig.5.Afewthings mightstandoutinthisimage.Thefirstoneisthat,forthese twocandidates,theco-solventcanbefoundconcentratedin bulkatthebottomofthedesorber,i.e.theco-solventdoesnot permeateallthewayuptothedistillate.Thishappenseven thoughfuranhasaboilingpointbelow35^◦C,whichisthetem- peratureofthecondenser.Thereasonisthattheco-solventis effectivelybeingwashedawaybytherichaminefedtothetop ofthedesorber,condensingintotheliquidphaseasitflows upwards.

ThesecondthingthatcanbenoticedonFig.5hastodo preciselywiththecondensationoftheco-solvent.Thisiswell evidencedbylookingattheconcentrationprofileofwaterin thecolumn(purplelines),speciallyforhigherco-solventflow rates.Thewatermolarfractioninthelaststageofthedesorber (S=5)ishigherthan thatinthereboiler(S=6),whichcan begraphicallyperceivedasabulgeinS=5.Thisisbecause theco-solventisbeinginjectedasvaportothereboilerand initiatingitscondensationpreciselyatthestagedirectlyabove it.Condensationisanexothermicprocess.Asaresult,thelast stageofthedesorberbecomeswarmerthanthereboileritself, whichleadstoincreasedwatervaporizationandabulgein itsconcentrationprofile.ThisisfurtherillustratedinFig.6, whichshowsthetemperatureprofilesinthedesorberwhen usingfuranasaco-solvent.(Noticethatthepointreferringto thestageS=0isnotshown,butthatwehavealreadyspecified thatthetemperatureinthecondenserisT=35^◦C.)

Thecondensationofco-solventandsubsequentformation ofatemperaturebulgecanbeseenaspotentiallybeneficialto theprocess,sinceitseeminglyinducesareductionoftheheat dutydemandsthatmustbesupplieddirectlytothereboiler.

However,thisimpressionismisleading.Firstly,thetempera- turebulgemightdefeatthepurposeoftheadditionofavolatile co-solvent,whichistokeepthetemperaturesinthecolumn

belowathresholdof120^◦Ctopreventaminedegradation.For example,theadditionofFCOS>0.1776kmol/hoffurantothe reboilermightpossibilitatetherecoveryofCO₂atabout340 kPawhilekeepingthereboilertemperatureat120^◦C,butthe stagerightabovethereboilerwillreachincreasinglyhigher temperatures,whichcreatesaclearbarriertohowmuchco- solventcanbeadded.Secondly,onecannotignoretheenergy required forseparating, heatingup and vaporizing the co- solventstreamwhencalculatingthetotalreboilerduties.For thisreason,insteadofpresentingthereboilerdutiessimply astheheatrequiredbythereboileralone,wehavedecided toaddtothisvaluetheenergyrequiredtoheatupaliquid co-solventstreamfrom25^◦Cupto120^◦Candtovaporizeit.

Thisisaconservativeapproach.Eveniftheliquidco-solvent is notrecoveredat25 ^◦C but atahigher temperature,one mustrememberthatthisco-solventhastobeseparatedby anunitaryprocesssuchasthroughtheuseofaseconddistil- lationcolumn,whichwilleffectivelyprobablyconsumemore energythanthesensibleheat+vaporizationheatthatisbeing consideredinthecurrentapproach.

Q=QR+QCOS=QR+

120^◦C 25^◦C

CP,COS·dT+H^vap_COS

· F_COS

D·y_CO2 (12)

Fig.7showssomeotherfeaturesoftheprocessoperating withaseriesoffuranasco-solvents.Ontheupper-leftcor- ner,onecanseehowincreasingco-solventflowratesallow forincreasingoperationalpressures,whicharehigherforthe morevolatileco-solvents.Whilefuranpossibilitatestherecov- ery ofCO2 ata maximumof230kPa, tetrahydrofuransets thelimitat225kPaand soforth.Theselimitingconditions are denoted bythestarsinFig. 7, whichmark thehighest FCOS foreach co-solventbeforetheestimatedtemperatures atanystageofthedesorbersurpass122^◦C.Theleastvolatile co-solvent, 3-methyltetrahydrofuran,activelydemands that the pressureofthedesorberisdecreasedsoastokeepthe reboiler temperature at120 ^◦C. On the upper-right corner, onecanseethatallco-solventsinduceanincreaseinenergy requirements followingtheapproachoutlinedinthe previ- ous paragraph. These shifting energyrequirements can be correlated tothechangesinrefluxratiosandboil-upratios seen on the bottom-left corner ofFig. 7. By promotingan overallincreaseindesorberpressureswhilefailingtoreach upwardstothedistillate,theadditionofthisseriesoffurans essentiallyactson thevaporizationofwateritself.Volatile co-solventssuchasfuranincreasethepressureofthedes- orbersothatlesswateriscondensedinthedistillate,which means lower reflux ratios. Conversely, boil-up ratios must increasetoaccountforthe circulationofthisnewaddition tothecolumn.Ontheotherhand,anon-volatileco-solvent suchas3-methyltetrahydrofuranallowsformorecondensa- tionofwaterthroughreductionoftheoperationalpressureof theprocess,increasingtherefluxratiosanddecreasingboil- upratios.Coupledwiththeheatrequiredforprovidingthis vaporizedco-solventstream,thenetresultoftheseeffectsis whatisobservedontheupper-rightcornerofFig.7.Finally, the bottom-rightcornerofFig.7showsthemolarfractions ofco-solventsinthedistillateandinthebottomproduct.As mentionedpreviously,theco-solventsintheseriesoffurans essentiallydonotreachthedistillate, meaningyCOS ≈0for thewholesetofsimulationsregardlessofsolventflowrates.

Fig.7–Resultsforsimulationswithseriesoffuransasco-solvents:pressureofthecolumn,totalheatduties,refluxratio andboil-upratio,co-solventconcentrationsintheproductstreams.Desorberwith5+2equilibriumstages.Therichsolvent molarflowrateis1kmol/h.Richloading˛=0.5,leanloading˛=0.2,temperatureatthereboilerof120^◦Candtemperature atthecondenserof35^◦C.ThestarsmarkthehighestevaluatedvalueofF_COSbeforethetemperatureatanystageofthe columnreaches122^◦C.

Asaresult,thebulkofco-solventaddedtothedesorbercomes outofthecolumnmixedwiththeleanaminestream,thus requiringasingleseparationprocessafterwards.

Finally,wemustmentionthat,thoughthetemperatureof thecondenserforallofoursimulationshasbeenkeptatTC= 35^◦C,theAppendixAofthisstudyprovidesananalysisofthe effectsofvaryingthisparameter.Ourresultsshowthat,when identifyingtheoverallbehaviorofthehigh-pressuredesorp- tionprocess,varyingT_Cisnotofhighestpriorityandcanbe ignoredforthesakeofsimplicity.

Thetakeawaysfromthisexercisewithaseriesoffuransas co-solventsare:

1 Thesolvents from this series are notvolatile enough to percolate the desorber all the way up to the distillate (Fig.5andFig.7),eventhoughtheboilingpointoffuran itselfiscalculatedat31.3^◦CwiththeparametersfromYaws (2003).Therefore,thewaythattheco-solventsofthisseries affectdesorberpressuresisbyaccumulatingatthebottom stagesofthecolumnwithoutcomingoutatthetop.

2 Theincreaseddesorberpressures inhibitwatervaporiza- tion(Fig.5).

3 Sincetheseco-solventsdonotreachthedistillateandact bydepressingthevaporizationofwater,thetemperatures

achievedinthecolumnmustbeoverallhigherthaninthe absenceofco-solvents(Fig.6).Thesetemperaturesincrease dueto(i)theco-solventstreamexothermicallycondensing inthecolumnand(ii)thewaterbeingunabletoendother- micallyevaporate.Iftherewasnoincreaseintemperatures inthedesorber,itisdoubtfulthatCO2wouldbesufficiently stripped:thedrivingforcetodesorbCO2mustbeprovided bysomethingelsenowthatlesswaterisbeingvaporized andnoco-solventiscomingtotheupperstages.Thissome- thingelseisheat.

4 Eventually,theadditionofco-solventformsatemperature bulklargeenoughsothatthewholepurposeofavoiding solventdegradationisdefeated(Fig.6).Therefore,thereisa captohowmuchco-solventofthiskindcanbeadded,and ofhowmuchincreaseinpressurecanbeattainedbythis methodology.

5 Astheco-solventcomesoutofthecolumnmixedwiththe leanamine (Fig.7), anewseparation stepisrequiredto removetheco-solventfrom thebottomstreamandkeep itinaclosedlooparoundthereboiler.

Overall,additionofaco-solventtypifiedbytheseriesof furansdoesnotseemtobeagoodideaforattaininghigher CO2deliverypressures.Thefactthatnoco-solventreaches

Fig.8–Vaporphasemolarfractionsofwater,amine,CO2andco-solventinadesorberwithisobutaneandnitrogen injection.Desorberwith5+2equilibriumstages.Therichsolventmolarflowrateis1kmol/h.Richloading˛=0.5,lean loading˛=0.2,temperatureatthereboilerof120^◦Candtemperatureatthecondenserof35^◦C.Theco-solventmolarflow ratesareprintedonthegraphs.

thedistillateimplyastrictcaponhowmuchpressurecanbe gainedbythismethodology.Withthisinmind,letusconsider thecaseforemployinghypervolatileco-solvents.

3.3. Desorberwithhypervolatileco-solvents

Byhypervolatileco-solvents,wemeanco-solventsthatarenot liquidsat25^◦Cand101.325kPa.Tobemoreprecise,inthis sectionwewillconsiderthecasesofdimethyletherandisobu- tane,whoseboilingpointsarerespectively−24.8^◦Cand−11.7

◦CfollowingtheAntoineparametersofYaws(2003).Addition- ally,purenitrogenhasbeenmodelledasaco-solventandits performanceispresentedinthis sectionaswell.Usingthe AntoineparametersprovidedbyYawsandSatyro(2015b),the boilingpointofnitrogenis−195.8^◦C.Thismeansthatnitro- gencannotbeseparatedfromCO2bycondensation,butrather thatCO₂itselfmustbecondensedoutofthedistillateincase thisco-solventisemployed.Inotherwords,nitrogenisnota practicalco-solvent,anditspresenceinthissectionismerely forillustrationpurposesasanexampleofacompoundwith veryhighvolatility.

Fig.8isverysimilartoFig.5,showingthevaporphasemolar fractionsofeach componentfora desorberoperating with isobutaneandwithnitrogenasco-solventsinjected atdis- tinctmolarflowrates.Thistime,however,onecanclearlysee

thattheco-solventpercolatesthewholecolumnandcomes outatthedistillate.Also,differentlyfrominthepreviousanal- ysis,thewaterconcentrationbulgeinS=5hasdisappeared.

Resultsfortheinjectionofdimethyletherarenotshownin Fig.8duetospacelimitations,andalsoduetothefactthat thecurvesobservedwiththisparticularco-solventfollowa trendinbetweenthoseobtainedbyinjectionofisobutaneand byinjectionofnitrogen.

Fig. 9 shows the temperature profiles in the desorber broughtbytheadditionofdimethyletherasco-solvent.As suggestedinthediscussionofFig.8,thetemperaturebulge causedbytheadditionofcompoundsintheseriesoffurans doesnotimmediatelyappearwhenemployinghypervolatile co-solvents.Indeed,itisseenthattheadditionofasmallFCOS

ofdimethyletherprovokesadecreaseintemperatureatthe bottomofthecolumnandanincreaseatthetop.Sincethecon- densationofco-solventisnolongerconstrainedtothebottom stages,thetemperaturesinthedesorberbecomemoreevenly distributed.Thisisnottosaythatasituationwillnotarise wherein atemperaturebulgecanbenoticed–itjustmight happen athigh co-solvent molarflow rates. Before that, a verynoticeable increaseintheoperationalpressuresofthe desorber canbeverifiedwiththeadditionofhypervolatile co-solvents.

Fig.9–Temperatureprofileswithdimethyletheras co-solvent.Desorberwith5+2equilibriumstages.Therich solventmolarflowrateis1kmol/h.Richloading˛=0.5, leanloading˛=0.2,temperatureatthereboilerof120^◦C andtemperatureatthecondenserof35^◦C.

Fig.10showsthesamesetoffeaturespreviouslydiscussed inFig.7,thoughtheresultsarenowmorecounter-intuitive.

Consider for example the plot in the upper-left corner of Fig.10.Thepressuresachievedbytheadditionofhypervolatile

co-solvents are clearly higher than those attained by the seriesoffurans,reaching500kPawithacomparativelysmall injectionofdimethylether.However,thistimethehigherpres- suresareprovidedbytheleastvolatileco-solvents:nitrogen ismorevolatilethandimethylether,whichismorevolatile thanisobutane,andyetitisisobutanetheco-solventthatis abletopressurizethedesorberthemost.Thekeytounder- standingthiscanbefoundinthebottom-leftcornerofFig.10.

Thehugemajorityofthenitrogeninjectedintothedesorber streamsupwardsandleavestogetherwiththedistillate,and thustherefluxratiodecreasessteeplyforhigherflowratesof nitrogen.Fordimethyletherandisobutane,therefluxratios are abitlarger,meaningthataparceloftheco-solventsis beingcondensedatthetopofthecolumnandbeingallowed torecirculate.Thisrecirculationcausestheconcentrationsof co-solventinbothliquidandvaporphasestobuildupmore thantheywouldotherwise.Itisthisbuildupthatallowsfor higher pressures to beachieved with isobutane than with nitrogen.

Inthebottom-rightcornerofFig.10,onecanseethatthe amountofdimethyletherandisobutaneleavingatthebot- tomfraction ofthedesorber together withthe lean amine isnotnegligibleatall,whereasthatofnitrogenapproaches 0.2%atbest.Infact,inthecaseofisobutane,thecondensa- tionoftheco-solventbecomessorelevantthataninflection in the molarfractionobtained inthe distillate isobserved

Fig.10–Resultsforsimulationswithhypervolatileco-solvents:pressureofthecolumn,totalheatduties,refluxratioand boil-upratio,co-solventconcentrationsintheproductstreams.Desorberwith5+2equilibriumstages.Therichsolvent molarflowrateis1kmol/h.Richloading˛=0.5,leanloading˛=0.2,temperatureatthereboilerof120^◦Candtemperature atthecondenserof35^◦C.ThestarsmarkthehighestevaluatedvalueofFCOSbeforethetemperatureatanystageofthe columnreaches122^◦C.

forincreasinglyhighF_COS (andincreasinglyhighpressures).

Athighco-solventflowratesF_COS,isobutanebehavesmuch likeoneofthefuranderivativesanalyzedinSection3.2:the co-solventbarely reaches thetop ofthe column, and both thereflux ratioRD and the co-solventconcentrationinthe distillateyCOS fallsteeply.Simultaneously,the formationof abulkofisobutaneatthebottomofthedesorbercreatesa clearthresholdatF_COS=0.0888kmolkmol/handp=390kPa, abovewhichthetemperaturesintheupperstagesofthecol- umnrise above 122 ^◦C.Thishas notbeen observed either fordimethyletherorfornitrogen,andthusonlythecurves referringtoisobutanearemarkedwithastarinFig.10.Inter- estingly,thus,thoughtheadditionofisobutanedeliversafast increaseindesorptionpressureswithcomparativelylowco- solvent flow rates, this increase is capped at p= 390 kPa, whereas nosuchcapisobserved fordimethyletheror for nitrogen.

Overall, hyper volatile co-solvents appear to be more promisingthan the chemicals exemplified bythe series of furans. Since the co-solvent percolates all the way up to the distillate, there is enough driving force to desorb CO2

evenwithoutlargeshiftsintemperature.Recirculationofthe co-solventthroughoutthe whole columnallowsforhigher pressures tobeachievedforsmallerflow ratesofadditive.

Finally,thoughsomeco-solventleavesthecolumnmixedwith the lean amine, its bulk can befound in the vapor distil- late together with CO2. Recoveryof the co-solvent can be performedbypressurization andcoolingofthedistillateor directlythroughchilling,dependingonthedifferenceofboil- ing points of vapor products. The boiling point of CO2 is

−78.5^◦Cat101.325kPaaccordingtotheparametersofYaws andSatyro(2015a),thusquitefarfromtheboilingpointsof dimethyletherandisobutane.

However,we must keepinmind that one ofthe objec- tivesofthisapproachistoreducethecompressiondutiesof theCO2captureplant.Letusconsidertheexampleofhigh- pressure desorptionwithdimethylether.On theupper-left cornerofFig. 10,oneseesthatthedistillate streamcanbe producedataround500kPawhen0.185kmol/hofdimethyl etherisinjectedintothereboiler.Thisisa165%increasefrom thedeliverypressureobtainedwithouttheadditionoftheco- solvent.Atthesametime,aparcelofthedimethyletherwill comeoutwiththedistillate.Intheaforementionedexample, themolarflowrateofvaporproductis360%higherthanthat whennotemployingtheco-solvent.Themethodtorecover thedimethyletherintheCO2streamwilldependontheavail- abilitiesatthe CO2 captureplantlocation.Ifthereare cold streamsthatcanbeused,perhapschillingisaproperalter- native fordimethylether condensation.Otherwise, cooling andcompressionmightbeabettersolution.Todiscussthese twoalternatives,wehaveemployedtheAntoineparameters ofCO2obtainedinYaws(2003)toperformflashcalculations ontheventproductofthedesorber.

Tocontinueontheexampleofhigh-pressuredesorption withdimethyletherasaco-solventdeliveringCO2at500kPa:

thedistillatecomesoutwith21.1%CO₂,78.3%dimethylether and0.6%waterat35^◦C.Ifonechoosestosimplycompress and cool down this product in acompression train, keep- ingthe temperatureat35 ^◦C,onewillseethat at1000kPa thereisstill55.5%dimethyletherinthevaporstream,then at1500kPathisvalueisreducedto33.7%,thenat2000kPa thedimethyletherconcentrationis22.7%andsoon.Itisonly ataround 5900kPaand35 ^◦Cthatthedimethylethercon-

centration inthe vaporstream fallsbelow1%, and thatof CO2consequentiallyreachesabove99%.Whatthismeansis that,althoughthepofcompressionwillbereducedwiththe useofdimethylether,theamountofgasbeingcompressed isincreasedsincenowitencompassesthedimethyletheras well.DependingontheconditionsoftheCO2captureplant, chillingtheproductcouldbemoreappropriate.Keepingacon- stantpressureof500kPa,thedimethyletherconcentration inthevapor fallsto49.5%at0^◦C,then to32.2%at−10 ^◦C andsoforth,finallyreachingbelow1%at−53^◦C.Otherwise, one could consider a combination ofchilling and pressur- ization:at1000kPathedimethyletherconcentrationinthe vaporproductreachesbelow1%at−36^◦C,andat1500kPa it isat−24^◦C.Wewillrefrainfrom goingintoodeepwith regardstodimethyletherrecoveryalternatives,butwewould liketostressoutthattheseconsiderations shouldbetaken intoaccountbeforeassertivelystatingwhetherhigh-pressure desorption with the injection of co-solvents is feasible or not.

Tosummarizetheresultsofthisexerciseonhypervolatile co-solvents:

1 Hypervolatileco-solvents(solventsthataregaseousat25^◦C and101.325kPa)candeliverhigherpressuresthanregular co-solventssuchasthoseofthefuranseries,inasmuchas theyareabletopercolatethewholedesorberandreachthe distillate(Fig.8andFig.10).

2 Withhypervolatileco-solvents,therecirculationofthesol- ventinsidethecolumnisafactorthatstronglyimpactsthe pressuresreachedinthedesorber.Withthatinmind,avery volatileco-solventmightbeworsethanalessvolatilesol- vent,sincethelattercanbeeasilyrecirculatedwhilethe formermightjustleavethedesorberinthedistillate(which, of course, depends on how the condenser is designed) (Fig.8andFig.10).

3 Asthe condensationof co-solventisa lesser issuewith hypervolatileco-solvents,temperaturebulgesarelargely avoided. Simultaneously, the fact that the co-solvent reachesthedistillatemeans itprovidesdrivingforcesfor thestrippingofCO₂.Thisraisesthethresholdforhowmuch co-solventcanbeaddedtothecolumn,andofhowmuch pressurecanbegainedwiththisaddition(Fig.9).Isobutane isafineexampleofaco-solventthatliesinthethreshold betweenhyper-volatileco-solventsandthoserepresented bytheseriesoffurans,behavinglikeeitherofthemdepend- ingontheoperationalconditionsofthedesorber.

4 Thehypervolatile co-solvent mostlikelyleaves the col- umnbothinthedistillate asinthebottomproduct, but mainlyinthedistillate.Thisco-solventinthedistillatecan berecoveredwithcompressionandcoolingorwithchilling.

Aspecificcase-by-caseanalysismustbeperformedtofind whichalternative,ifany,enablestheuseofhigh-pressure desorption.Theco-solventinthebottomproductmustbe recoveredwithasecondaryseparationprocess.

Thelastpointisperhapsoneofthemostimportantcon- clusions ofthiswork.Ifit istruethathigherpressures are achievedwhentheco-solventisallowedtoreachthedistillate butalsocondensesandrecirculateinsidethedesorber,then itisinevitablethataproperco-solventforhigh-pressuredes- orptionwillcomeoutfractionedbetweenthevaporandliquid productsofthestripper.Thenonewillrequiretwoextrasepa- rationsteps,notmerelyone,inordertorecovertheco-solvent.

4. Conclusion

High-pressuredesorptionhasbeenevaluatedasanalterna- tiveforproducingCO2athigherpressureswhilstkeepingthe maximumsolventtemperaturebelow120^◦C,thusavoiding aminedegradation.Ashortcutmethodologyhasbeendevel- opedto quicklyevaluateaseries ofcandidate co-solvents, though this methodology is handicapped by the fact that itignores theactualconditions ofthe desorberasawhole (refluxand boil-up ratios,forexample). In due course,the modellingandsimulationofthestripperhasbeenperformed withtheinjectionofaseriesofco-solventsoflowtomoderate volatilityandthenwithaseriesofhypervolatileco-solvents.

Lowvolatileco-solventssuchas2-methyltetrahydrofuranand tetrahydrofurandonotseemtoprovidepressureshighenough tojustifytheirutilization.Simultaneously,theircondensation inthedesorberpossiblycreatesatemperaturebulgethatlim- itshowmuchpressurecanbeattainedthroughthisapproach.

Alternativehypervolatileco-solventscondenselessandper- colatethedesorberallthewayuptothedistillate,providing higherpressuregainsandavoidingtheformationoftemper- aturebulges.Thistrendisotherwisesubvertedinthecaseof nitrogen,whichissovolatilethatit simplyleavesthedes- orberwithlittle tonorecirculation,building comparatively small amounts of pressure.As such, there seems to be a limitedvolatilityrangerequiredfortheproperdesignofhigh- pressuredesorptionwithco-solvents.Nevertheless,thisrange isdefinedbythespecificationsinwhichthedesorbercolumn isoperated,specificationswhichhavebeenproposedinthis studyonlyassurrogatesforthesakeofanalyzingtrendsand identifyingpatterns.

Theuse ofvolatile co-solvent injectionas a means for recoveringCO2athigherpressuresseemstobetheoretically feasible and a promisingalternative process configuration.

However,itisalsosubjecttoaverycarefuloptimizationprob- lem.Asuitableco-solventmustmakeacompromisebetween beingvolatileenoughtodeliverhighpressures andnottoo volatilesothatitwillstillbeabletorecirculate inthedes- orber.Additionally,itmusthavesuchpropertiessothatitis easilyrecoverablefrombothliquidandvaporproductswith- outincurringintoomanyextraoperationalandcapitalcosts.

Wearehopefulthatthepreliminaryanalysisperformedinthis workcanshedlightonthetrade-offsinherenttothisnovel high-pressuredesorptionprocess.

Symbol Units Meaning

Latinletters

B kmol/h Bottomproduct

D kmol/h Distillate

F kmol/h Feedmolarflowrate

fCOS mol

co-solvent/mol water

Co-solventfraction

k1,k2,k3 — Parametersofsoftmodel

L kmol/h Liquidmolarflowrate

N — Numberofinnerstagesof

desorber

n_i mol Numberofmolsofi

p kPa Totalpressure

p_i kPa Partialpressureofi

Q MJ/kgCO₂ Totalregenerationduty QC MJ/kgCO2 Condenserduty QCOS MJ/kgCO2 Dutyforrecoveryof

co-solvent QR MJ/kgCO₂ Reboilerduty

RB — Boil-upratio

RD — Refluxratio

S — Stageofthedesorber

T K Temperature

V kmol/h Vapormolarflowrate

xi — Molarfractionofiinliquid

y_i — Molarfractionofiinvapor

Greekletters

˛ molCO2/mol

amine

Loading

H kJ/mol Enthalpyofphasechange

1,2,3 — Degreesofadvancement Subscripts

CO2 ReferringtoCO₂

COS Referringtothevolatileco-solvent H2O Referringtowater

MEA Referringtomonoethanolamine FEED Referringtotherichaminefeed Superscripts

abs Referringtoabsorption

app ReferringtotheaminesolventwithoutCO2

eff Referringtotheaminesolventonce reactedwithCO2

fresh Referringtotheaminesolventwithout CO2norco-solvent

sat Referringtosaturation vap Referringtovaporization

Declaration of interests

The authors declare that they have no known competing financialinterestsorpersonalrelationshipsthatcouldhave appearedtoinfluencetheworkreportedinthispaper.

Acknowledgements

ThisresearchwasfundedbytheFacultyofNaturalSciencesof theNorwegianUniversityofScienceandTechnology(NTNU).

Appendix A

TheAntoineequationforcalculatingsaturationpressuresis:

log₁₀

=Ai− B_i C_i+T−273.15

WherethetemperatureTissuppliedinKandthesaturation pressurep^satisreturnedinkPa.ThecoefficientsA,BandCfor differentcomponentsareshowninTableA1.Thedimensions

TableA1–Antoinecoefficientsforthecalculationofvaporpressure.

Component A B C Source

Water 8.05573 1723.6425 233.08 (1)

Monoethanolamine 7.44237 1560.9667 171.200 (2)

Carbondioxide 7.58828 861.82 271.883 (3)

Furan 7.13277 1145.36 238.023 (3)

Tetrahydrofuran 7.10537 1256.68 232.621 (3)

2-methyltetrahydrofuran 7.13891 1339.48 234.353 (3)

3-methyltetrahydrofuran 6.99166 1430.57 210 (3)

Dimethylether 7.19658 984.579 252.976 (3)

Isobutane 6.93388 953.92 247.077 (3)

Nitrogen 6.72531 285.5727 270.09 (1)

Sources:1=YawsandSatyro(2015b),2=YawsandSatyro(2015a),3=Yaws(2003).

TableA2–Polynomialcoefficientsforthecalculationofgasheatcapacity.

Component a b×10³ c×10⁵ d×10⁹ e×10¹¹

Water 33.174 –3.2464 1.7437 –5.9796 —

Monoethanolamine 33.174 –3.2464 –31.976 158.3 –3.2344

Carbondioxide 27.437 42.32 –1.9555 3.997 –0.029872

Furan –13.779 334.89 –22.273 –69.36 –0.81619

Tetrahydrofuran 32.887 24.554 60.226 –623.8 18.528

2-methyltetrahydrofuran –15.65 607.52 –36.17 79.1 —

3-methyltetrahydrofuran –15.65 607.52 –36.17 79.1 —

Dimethylether 34.668 70.293 16.53 –176.7 4.9313

Isobutane 6.772 341.47 –10.271 –36.85 2.0429

Nitrogen 29.342 –3.5395 1.0076 –4.3116 0.025935

Source:Yaws(2003),exceptnitrogen,whichcomesfromCoker(2007).

ofthesecoefficientsarerespectivelynone(Aisadimensional), KandK.

Theheatcapacityofasinglecomponentinthegasphase isgivenby:

C_P,i=a_i+b_i·T+c_i·T²+d_i·T³+e_i·T⁴

WherethetemperatureTissuppliedinKandtheheatcapacity CPisreturnedinJ/mol·K.Thepolynomialcoefficientsa,b,c, dandefordifferentcomponentsareshowninTableA2.The dimensionsofthesepolynomialcoefficientsarerespectively J/mol·K,J/mol·K²,J/mol·K³,J/mol·K⁴,andJ/mol·K⁵.

TableA3–OrganiccompoundsandthetotalpressureattainedinthereboilerforMEA30%wt.when␣=0.2,f_COS=0.1and T=120^◦C.CaseA:co-solventcanberecoveredasliquidat25^◦Cand101.325kPa.

Structure Name Reboilerpressure/kPa Hazards

2,3-butadien-1-ol 959.6

Vinylformate 753.9 couldnotfindinformation

3-methoxy-1-propene 394.8

1,2-dimethylcyclopropane 374.3/343.7 unstable

Dimethylacetylene 369.3

Divinylether 357.9 unstable

3-methyl-1-butyne 350.5

Propyleneoxide 347.7/322.6/319.8

Methylvinylketone 343.0