Salt precipitation due to supercritical gas injection: I. Capillary-driven flow in unimodal sandstone

(1)

ContentslistsavailableatScienceDirect

International

Journal

of

Greenhouse

Gas

Control

jo u r n al hom e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / i j g g c

Salt

precipitation

due

to

supercritical

gas

injection:

I. Capillary-driven

ﬂow

in

unimodal

sandstone

H. Ott

a,∗

_,

_S.M.

_Roels

b

_,

_K.

_de

_Kloe

a

a_Shell_Global_Solutions_{International}_B.V.,_Kessler_Park_1,₂₂₈₈_GS_Rijswijk,_The_Netherlands b_Department_of_{Geotechnology,}_Delft_University_of_Technology,₂₆₂₈_CN_Delft,_The_Netherlands

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received4October2014

Receivedinrevisedform5December2014 Accepted6January2015 Availableonlinexxx Keywords: CO2storage Dry-out Precipitation Counter-currentﬂow Micro-CTimaging Sandstone

a

b

s

t

r

a

c

t

Dryingandsaltprecipitationingeologicalformationscanhaveseriousconsequencesforupstream oper-ationsintermsofinjectivityandproductivity.HereweinvestigatetheconsequencesofsupercriticalCO2

injectioninsandstones.ThereportedﬁndingsaredirectlyrelevantforCO2sequestrationandacid–gas

injectionoperations,butmightalsobeofinteresttoabroadercommunitydealingwithdryingand capillaryphenomena.

ByinjectingdrysupercriticalCO2intobrine-saturatedsandstone,weinvestigatethedryingprocess

andtheassociatedprecipitationofsaltsinacapillary-pressure-dominatedflowregime.Precipitation patternswererecordedduringthedryingprocessbymeansof␮CTscanning.Theexperimentalresultsand numericalsimulationsshowthatunderacriticalflowratesaltprecipitateswithaninhomogeneousspatial distributionbecauseofbrineandsolutesbeingtransportedincounter-currentflowupstreamwhere salteventuallyprecipitates.Asubstantialimpairmentoftheabsolutepermeabilityhasbeenfound,but despitehighlocalsaltaccumulation,theeffectiveCO2permeabilityincreasedduringallexperiments.This

phenomenonisaresultoftheobservedmicroscopicprecipitationpatternandeventuallytheresulting K()relationship.

Theﬁndingsinthispaperarerelatedtounimodalsandstone.Inacompanionpaper(Ottetal.,2014) wepresentdataonthedistinctlydifferentconsequencesofsaltprecipitationindual-ormulti-porosity rocks.

1. Introduction

Dryingofporousmediaisanimportanttopicinmany indus-trialprocesses,insoilscienceandforupstreamoperationssuch asgasinjectionandproductionintoandfromgeological forma-tion.Dryingofsalineformationswillcauseprecipitation ofsalt initiallydissolvedinthebrine.Thiscannegativelyaffectthe perfor-manceofinjectionandproductionwellsandcanevenleadtowell clogging,whichisaseriousriskforsuchoperations.Inthispaper weconsiderlarge-scalegeologicalstorageofCO2,originatingfrom anthropogenicsourceslikefossil-fueledpowerplantsor contami-natedgasproduction,inordertoreduceCO2emissions.Deepsaline aquifersanddepletedoilandgasﬁeldsarepotentialsubsurface depositsforthatpurpose(IPCC,2005;BachuandGunter,2004).

∗ correspondingauthor.Tel.:+31704473043. E-mailaddress:Holger.Ott@Shell.com(H.Ott).

Ifdryorunder-saturated,supercritical(SC)CO2isinjectedinto water-bearinggeologicalformationslikesalineaquifers,wateris removedeitherbyviscousdisplacementoftheaqueousphaseor byevaporation/dissolutionofwaterinCO2andsubsequent advec-tionintheinjectedCO2-richphase.Bothmechanismsactinparallel, butwhileadvectionoftheaqueousphasedecreaseswith increas-ing CO2 saturation(diminished mobility), evaporationbecomes increasinglyimportantastheaqueousphasebecomesimmobile. Belowresidualwatersaturation,onlyevaporationtakesplaceand theformationdriesoutifnoadditionalsourceofwaterisavailable. Ifwaterevaporates,thesaltsoriginallypresentinthewaterareleft behind.Inhighlysalineformations,theamountofsaltthat poten-tiallyprecipitatesperunitvolumecanbequitesubstantial.The volumesdependonbrinesalinity,andthetransportofsolutesand waterinthereservoir.Sinceﬂuidsaturationsandﬂowratesclose tothewellborecoveralargerangeasfunctionsofspaceandtime, therearenoeasyanswerstothequestionswhether,whereandhow saltprecipitatesandhowprecipitationaffectsinjectivity.The ques-tionsthatneedtobeaddressedareaboutthemechanismsofsolute http://dx.doi.org/10.1016/j.ijggc.2015.01.005

(2)

ARTICLE IN PRESS

G Model

IJGGC-1398; No.ofPages9

2 H.Ottetal./InternationalJournalofGreenhouseGasControlxxx(2015)xxx–xxx

transportonamacroscopicscalethatdeterminethemacroscopic distributionofsalt,andthesaltdistributiononaporescalethat determineshowthepermeabilityisaffectedasfunctionofporosity reduction.

Eventhoughsaltprecipitationinthevicinityofgasproduction wellsaregenerallyconsideredasissue,therearenotmany stud-iespublishedreferringtowelldata.Thefrequentlycitedpaperis thatofKleinitzetal.(2001).Probablyduetothelimitednumberof industrial-scaleCO2sequestrationprojects,therearenowelldata availablereferringtosaltprecipitationduringCO2injection.

Thelackofwellstudiesrendersaproperproblemstatement difficultandwerelyonnumericalsimulationsandanalytical mod-elsthatareavailableinliterature.Here,weonlyaimtohighlight afewkeystudiesonprecipitationduetoCO2 injectionrelating totheproblemdiscussedinthispaper.Inaseriesofpublications, Pruessetal.presentednumericalsimulationsofCO2injectionin salineaquifers,investigatingthefundamentalaspectsofformation dry-outandsaltprecipitation(PruessandGarcía,2002;Pruessand Müller,2009).Thesimulationswereperformedwithasingle injec-tionwellinidealized1Dand2Dradialgeometries.Theauthors observedthatprecipitationoccursonlyinanarrowdry-outzone confinedtoafewmetersaroundtheinjectionwell.Thesolidsalt saturationin thiszone hasbeenfoundtobeconstant indepen-dentoftheinjectionrate.2Dradialsimulationswerecarriedout toexploregravity effects. Incontrast tothe1D-radialscenario, gravityin combination withcapillary-driven flow lead tomore heterogeneousprecipitation,withamaximumobservedsolid-salt saturationofmorethan20%.

Giorgisetal.(2007)performedﬁeld-scalesimulationinradial geometry. The authors found that the amount of precipitate dependsonbrinemobilityandcanbehighifthereisa capillary-gradientdrivenbrineﬂowinthedirectionofthewellbore.The authorshavefurthershownthattheinjectionrateisanimportant factorincontrollingprecipitationprocessandinavoidingor allow-ingcompletecloggingoftheformation.Intheirsimulations,solid saltsaturationsoflocallymorethan60%havebeenreached.

However,field-scalesimulationsrequireinputonflowphysics andthermodynamicssuchastheK()relationshipandthemass transferratesbetweenthefluidphases.Aswewillshowinthis paperandinOttetal.(2014),theseparametersareofmicroscopic originandneedtobedeterminedbylaboratoryexperiments.The qualityoftheinputiscriticalfor reliablesimulationsasseveral studiessuggestthata modestchangeinporosity mightleadto a seriousreduction in permeability. Therespectiveliterature is diverseinmechanismsandtherearenotmanystudiesdealingwith fluid-transportinducedporositychanges. Thediscussed mecha-nismsrangefromtheporosityvariationduetolithology/rocktype (Papeetal.,1999;EhrenbergandNadeau,2005)viamechanical compaction(Wyble,1958;Schutjensetal.,2004)tosilica disso-lution and precipitation in geothermalsystems in single-phase flow(Xuetal.,2004b)todryingprocesses,i.e.intwo-phaseflow asdiscussed inthe presentpaper. Evenifusually described by power laws, it cannot be expected that the K() relationships resultingfromdifferentmechanismsarecomparable–i.e. process-independent–andgenerallyapplicable.

Thereareonlyafewstudiesonﬂow-throughdryingavailable, which are relevant for CO2 storage.Zuluaga et al. investigated vaporizationandsaltprecipitationinsandpacksand sandstone forgasproductionwells (Zuluaga andMonsalve,2001; Zuluaga etal., 2001).AttheGHGT-10 in Amsterdam2010, two experi-mental studies ondry CO2 injectionhave been presented. The experimentshavebeenperformedinsandstoneinrealisticstorage conditionsaddressingcapillary-drivensolutetransport,the condi-tionofcountercurrentﬂow(Ottetal.,2011b),andapermeability porosityrelationship(Baccietal.,2011).Recently,Peyssonaetal. (2014)andAndreetal.(2014)investigatedthedryingprocessby

nitrogeninjectioninsandstone.Thedatahavebeenusedto bench-markanumericalsimulationtoolforfield-scalemodelingofCO2 injection.Ottetal.(2011b)pointedoutthatmodelingof vaporiza-tionbyanequilibriumapproachisnotsufficienttodescribecore floodexperiments.Roelsetal.(2014)performedcoreflood exper-imentsandsucceededinthedescriptionofthesaturationprofiles byakineticapproach.

Inthispaperweshowresultsofcore-floodexperimentsthat werepresentedattheSCAconferenceinHalifaxandattheGHGT10 inAmsterdamin2010(Ottetal.,2010,2011b).Theexperiments wereperformedtoinvestigatethedryingprocessandtheimpact ofsaltprecipitationonflow,i.e.thesaltdistributionandtheK() relationship.Forthis weinjecteddry SCCO2 in brine-saturated siliciclasticsandstone.Theexperimentswerecarriedoutatflow ratesrealisticfornear-well-boreflowandatrealistic thermody-namicconditions.Duringinjection,spatialandtimeevolutionof saturationchangesweremonitoredbymeansofmicrocomputed tomography(␮CT).Theresultsinthispaperarebasedona num-berofexperimentsshowingprecipitationprofileswithdifferent degreesofheterogeneityonamacroscopicscale.For quantifica-tionwediscusstwooftheseexperimentsshowingthelargestand thesmallestspatialvariationofsaltsaturationafterdry-out.We refertothecasesastheheterogeneousandthehomogeneouscase, respectively.Weexplainthemechanismsthatleadtotheobserved heterogeneousdistributionoftheprecipitatebymeansof numer-icalsimulations.

Inadditiontothesaturationprofiles,changesinabsolute per-meabilityandeffectiveCO2 permeabilityweremonitored.From thisdata,weextractthepermeability/porosityrelationship(K()). Weexplaintheobservedmildpermeabilityreductionbythe micro-scopicdistributionofthesaltwithrespecttotheobservedCO2flow channels.Furthermore,wediscusstheresultsintermsof precipi-tationinsingle-phaseandtwo-phaseflowsituations.

2. Materialsandmethods

Theexperimentswerecarriedoutinacorefloodsetupdesigned forfloodingwithvolatileandreactivefluidsassketchedinFig.1.A detaileddescriptionoftheunitcanbefoundelsewhere(Ottetal., 2012).Inthefollowing,onlyabriefdescriptionoftheelementswill begiventhatareofrelevancefortheexperimentsherepresented. Theflow experiments were performedin verticalgeometry, withfluidsbeinginjectedfromtoptobottom.Thesampleswere embeddedinpolycarbonateandplacedinacarbon-fiberbasedcore holder–botharematerialswithlowX-rayattenuationcoefficients. Thecoreholderisplacedina␮CTscannerforinsitu3Dimaging oftherock-fluidsystem.CTimagingallowstodeterminefluid sat-urationsandchangesoftherockmatrixduetosaltprecipitation. Theunitisequippedwithtwofeedsectionsforliquidsand lique-fiedgasinjection.TheCO2feedpumpwasheldat3◦Cduringthe experiments–liquidCO2wasinjectedandheatedto experimen-taltemperatureandtotherespectiveSCstateintheinjectionlines. Thedensitydifferencehasbeentakenintoaccountfortheindicated flowrates.Fromflowratesandthedifferentialpressure measure-ment(P),theabsolute(K)andeffective permeability(K×krel) werederivedonline.

TheexperimentswereperformedonBereasandstonewithan averagepermeabilityof 500mDand 22%porosity. Thesamples weredrilledfromthesame blockandwere smallin cross sec-tionandvolume(1cmøand5cmlength)toobtainrepresentative ﬂowratesandtoreducetheexperimentaltimetocomplete dry-out.ThemineralogyofBereaisdominatedbyquartzwithsome K-feldspar,kaolinite,andminoramountsofotherclaymineralsas determinedbyeSEM/EDX.Therocksampleswerepre-saturated withNaCl-basedhigh-salinitybrine:20wt%NaCland2wt%CsCl.

(3)

Fig.1. Schematicoftheexperimentalsetup.MoredetailsareprovidedinOttetal. (2012).

TheCsClwasaddedasacontrastagentasitsX-rayabsorption coef-ﬁcientishigh,leadingtoahighX-rayabsorptioncontrastin␮CT betweentheaqueousandtheCO2-richphase.Theinjected CO2 wasofhighpurityandessentiallydry(stronglyunder-saturated withrespecttowater),notleastbecauseofthestrongdifference betweenwatersaturationlimitsatconditionsinthelq.-CO2 cylin-derandatexperimentalconditionof100barpressureand45◦C temperature.

We observe the spatial distribution of the precipitated-salt phasebymeansof␮CTimaging.Computertomographyisbased onX-rayabsorptiondeterminedbymaterial-specificlinear atten-uation coefficients (), which are directly represented as gray valuesofthevoxelsinthereconstructedimage(Wellingtonand Vinegar,1987;VinegarandWellington,1987).Forsaturation cal-culationswemadeuseoftabulatedmass-absorptioncoefficients /(Hubbell,1969),fluiddensities(PruessandSpycher,2007; SpanandWagner,1996),andthedensityofthesolid-saltphase (Lide,2003)./fortheinjectedandproducedfluidsandtheir con-stituentsareshowninFig.2andphasedensitiesarelistedinTable1. Becauseofthesensitivityofsaturationcalculationstothe X-ray-contrastagent,wemeasuredrelativeabsorptioncoefficientsofCsCl solutionsasafunctionofCsClconcentration.Thecurveisshown inthelowerpanelofFig.2;at2wt%CsCl, themass-absorption coefficientisabout1.6timeslargerthanthatofpurewater.The CsClcalibrationcurve wasalsousedtodeterminetheeffective photonenergyofthe␮CTphotonsource(heff=63keV)inorder toderivethemassabsorptioncoefficientsusedforthesaturation

Fig.2. Top:X-raymass-absorptioncoefficientsoftherelevantfluidsandsaltsas functionofphotonenergy.Theverticallineindicatesthecalibratedeffectivephoton energythathasbeenusedtocalculatethesaltsaturationbyEq.(2).Bottom:relative absorptioncoefficientsofthebrinephaseasfunctionofCsClconcentration.Thedata hasbeenusedtodeterminetheeffectivephotonenergyh=63keV.

calculations.Thederivedeffectivemassabsorptioncoefﬁcientsare listedinTable1.

DuetothehighersolubilitylimitofCsCl,weexpectthatsalt pre-cipitatesinaratiodifferentfromtheinitialratiomNaCl/mCsCl=10, becauseCsClmightstaylongerinsolutionandmightbe prefer-entiallyremovedbythedisplacementprocess.ToestimateCsCl depletion,weperformedsimulations(notshown)withthe exper-imental brine composition.The simulations wereperformedas outlinedinSection4,butusingToughReact(Xuetal.,2004a)as simulation tool. NaCl and CsCl precipitations were handled in equilibriumusingtheToughReactdatabaseforNaClandthe CsCl-solubilitydatareportedinLide(2003).Thesimulationsresultin mNaCl/mCsCl=12.3and arespectively correctedscalingfactorfor saturationconversionaccordingtoEq.(2).

3. Longitudinalprecipitationproﬁle

Foreachexperimentadryandcleanrocksamplewasmounted into the core holder and subsequently pressurized and heated under N2 ﬂow to 45(±1)◦_C/100_bar _(downstream _side), corre-spondingtothethermodynamicconditionsofasalineaquiferat adepthofabout1000m.Toquantitativelydeterminesaturations, referencescansat100%CO2saturationand100%brinesaturation wererecorded.Forthatpurpose,nitrogenwasdisplacedbySCCO2 (miscibledisplacement)until100%CO2 saturationwasreached, ascheckedbythedensityoftheproducedﬂuid.Subsequently,the corewasslowlydepressurizedandevacuatedtoallowbrine satura-tionwithouttrappingCO2.Thecorewasthensaturatedwithbrine andsubsequentlypressurized.Allthereferencescansweretaken atexperimentaltemperatureandpressure.Duringtheexperiment, drySCCO2wasinjectedataconstantrateintothebrine-saturated core.Theinjectionrateatexperimentalconditionwas2.2ml/min

(4)

ARTICLE IN PRESS

G Model

4 H.Ottetal./InternationalJournalofGreenhouseGasControlxxx(2015)xxx–xxx Table1

Densitiesat100barand45◦_C_and_{mass-attenuation}_{coefﬁcients}_for_converting_saturation_proﬁles_according_to_Eq.₍₂₎_;_*₌_dry,_**₌_water_saturated,_†₌_with,_and_††₌_without CsCl,‡=inthepresenceofCO2‡‡=beforeCO2-breakthrough.§:thedeviationfromtheliteraturevaluecanbeexplainedbya≈2Klowertemperatureintheexternaldensity meter. CO2(kg/m 3₎ brine(kg/m3) salt(kg/m3) Measured 585**_→₅₆₅*, ₁₁₄₀†,‡‡_(±2) _– Literature 499*_,₅₁₇*,** ₁₁₄₀††_,₁₁₇₂††,‡ ₂₁₇₀_(NaCl),₃₉₈₈_(CsCl), (SpycherandPruess,2005) (SpycherandPruess,2005) (Lide,2003)

2264(mNaCl/mCsCl=10)

Usedforsaturationconversion 500 1140 2247(mNaCl/mCsCl=12.3)

(/)CO2(cm

2_/g) _(/)

brine(cm2/g) (/)salt(cm2/g)

0.85(mNaCl/mCsCl=10)

Usedforsaturationconversion 0.185 0.33 0.75(mNaCl/mCsCl=12.3)

(later4.4ml/min)correspondingtorealisticnear-well-bore flow rates.Duringthefirst40minbrineandCO2wereproducedand,in thefollowingperiod,onlyaCO2-richphasewithnofurtherbrine production.Atthatpoint,substantialdryingbyevaporationstarted, withadifferentialfluidpressureofabout600mbar,decreasingto alowerfinalvalueof460mbarafterabout8.5hofCO2floodingas laterdiscussedinFig.8.

Fig.3 showsthe integrated ␮CT-response profilesalong the flowdirectionatdifferentexperimentaltimesteps.Theprofiles arealreadyconvertedtosaturationsbyeliminatingtherockmatrix. TheCO2saturationisrepresentedwithascalerangingfrom100% brinesaturation(_SCO₂₌0)to100%CO2saturation(SCO2=1).The

saturationshavebeencalculatedfromintegratedCTproﬁlesI(t)by: SCO2 =

Ibrine−I(t) Ibrine−ICO2

(1)

whereas Ibrine and ICO2 represent thedensity proﬁles of the

referencescansat100%brineand100% CO2 saturation, respec-tively.TheﬁrstCTscanwasstartedafter0.4hofCO2ﬂooding.The

Fig.3.␮CTdatarecordedduringtheexperiment.Top:Timeseriesofnormalized differenceimagesprojectedontotheverticalsampleaxisinﬂowdirection.The proﬁlesrepresentchangesofathree-phasesystem:brine/CO2/salt.Thesaturation scalesontheleftandrightcorrespondtotherespectivetwo-phasesystematthe beginningoftheexperimentandattheend:CO2/brineandCO2/salt,respectively.

saturationproﬁleisﬂat,withanaverageCO2saturationofabout0.5 (bluesymbols).Thedisplacementwasstilladvection-dominated. Saltprecipitationdue toevaporationcanbeignored andsothe systemwasinatwo-phaseregime(CO2-brine),forwhichSCO2isa

goodscale.

SubsequentlytheX-rayabsorptiondecreaseswithtime,inline withanincreaseinCO2 saturation.Ifsaltdidnotprecipitate,flat densityprofileswouldbeexpected,determinedbytheadvection andevaporationofthebrinephase.However,whilethelightest componentinthisexperiment(CO2)wasinjected,adipoccurred atapositionofabout2.5mm,correspondingtoadensityincrease– saltprecipitates!Thedipgrewforabout9.5h.After10h(red sym-bols)theshapeoftheprofile,includingallfeatures,didnotchange furtherforanother6hofCO2flooding.Thisisaclearindication thatthesampleisdry–nowaterevaporatesandnosaltis precip-itatinganymore.Weassumethatatthatpointintime,therock samplecontainedsolidsaltandCO2,whichagainisatwo-phase system.ThisallowstheCTresponsefortheredprofile(drystate) toberescaledtosolidsaltsaturationby:

Ssalt=(1−SCO2)·

brine−CO2

salt−CO2

, (2)

where the s denote the respective attenuation coefﬁcients, derived from the mass-attenuation coefﬁcients (/) and the respectivephasedensities,listedinTable1,by=(/).

The mean value of solid-salt saturation was determined as 4.1(_±0.2)%.1_However,_locally_the_saturation_is_as_high_as_18%._The

meanvalueof4.1%correspondstoevaporationof41(±2)%ofthe totalwater−theimmobilewaterfraction.Fromthis,theaverage degreeof water saturationin theproduced CO2 hasbeen esti-matedtobeabout35(±2)%usingthesaturationlimitofwaterin CO2(≈0.00145wt.%,(PruessandSpycher,2007;SpanandWagner, 1996)),theinjectionrateandtheexperimentaltimetodry-out. Fromthisweestimatethelengthscaleoftheevaporationzoneto beintheorderof0.3–1m,whichevidentlyiscontrolledby evapo-rationkineticsandtheﬂowrateoftheCO2-richphase.Anumerical estimationofthelengthoftheevaporationzonewillbegivenin Section5.

Theobservedsaltaccumulation,whichislocallymuchhigher thanthesaltoriginallypresentinthebrineinthesamerespective volume,isexplainedbyacapillary-drivencounter-currentﬂow, asdiscussedfurtherinthenextsection.However,afterdoubling theﬂowrateto4.4ml/min,ahomogeneousprecipitationpattern hasbeenobserved.Bothprecipitationpatternsareshowninthe lowerpanelofFig.3.Anexplanationofthiseffectwillbegivenby

1_The_saturation_{determination}_is_mainly_sensitive_to_the_uncertainty_in_the_dopant concentration,whichmightbeaffectedbydepletionasdiscussedinSection2.The givenerrorsresultformtheestimatederroroftheNaCl/CsClratioof(12.3)±1.

(5)

numericalsimulationsinthenextsection.Thereitturnsoutthat thereisacriticalflowrateabovewhichsaltprecipitates homoge-neously.Itneedstobementioned,thatwedidnotobserveawell definedcriticalflowrate,whichis–aswebelieve–aresultofthe (relativelysmall)sampletosamplevariationofcapillary proper-ties.Whatwecanstateatthispointisthatthecriticalvelocityfor thegivenrocktypeandthermodynamicconditionisintherange oftheflowratesreportedhere.

4. Numericalmodelingofthelongitudinalproﬁle

Foranunderstandingofthefundamentalmechanismbehind theobservedprecipitationpatterns,weperformednumerical sim-ulations with TOUGH2 (Pruess and Spycher, 2007), which is a multiphase/multi-componentreservoirsimulator.Incombination withthefluidproperty moduleECO2Nthatdescribesthe ther-modynamicsofH2O/NaCl/CO2systems,TOUGH2hasbeenusedto modelCO2injectioninsalineaquifers(PruessandSpycher,2007). Thepresent experimentsweremodeledin asimple vertical1D geometryusingtheexperimentaldimensions.Thefirstgridblock wasassignedastheinjectionpointataconstantCO2-injectionrate. Theconstantpressureboundaryconditionwasrealizedbysetting thevolumefactorofthelastgridblocktoinfinite.Therock-fluid propertiesK,k_relandpCwerederivedfromexperimentsperformed onthesamerocktypewiththesamefluidconfiguration(Perrinand Benson,2010;Bergetal.,2013).2

Theupperpanelof Fig.4shows longitudinalCO2-saturation profilesprofile(inthedirectionofinjection)simulatedwithaCO2 -injectionrateof3×10−5kg/sforseveraltimesteps(blacklines). Atthegiveninjectionrate,theshockfrontpassesthecorewithin seconds,followedbyacontinuoussaturationchangeafter break-through.Afterabout0.3h,adipintheCO2saturationclosetothe injectionpointoccurs.Thisisattributedtoasaturationchangedue tosaltprecipitationasdisplayedinthelowerpanel.With advanc-ingtime,aprecipitationfrontslowlymovesfromtheinjectionpoint downstream,accumulatingsaltonitsway.After3.3hofCO2 flood-ing,thesolid-saltsaturationexceeds0.9andthesimulationstops; inthenexttimestepthesaltsaturationwouldreach1. Permeabil-itychangesarenottakenintoaccount,butablockageofporespace willobviouslyleadtoclogging.Atthatpointmoresaltis precipi-tatedthanwasoriginallypresentintherespectivebrinevolume, whichrequirestransportofsaltintheupstreamdirectiontothe pointofdry-out,asalsoobservedexperimentally.

InthesamepanelsinFig.4,simulationsathigherandlower injectionratesaredisplayedintheirrespectivefinalstatesafter dry-out.Athigherinjectionratesthelocationofprecipitationshifts awayfromtheinjectionpointtowardalargersaturationgradient (asaresultofthecapillaryendeffect).Apparently,iftheviscose forceincreases,astrongersaturationgradientisneededtoreach theconditionthatleadstolocalprecipitation.Atlowerflowrates, backflowseemstobeevenstronger,withthesaltprecipitating directlyattheinletwiththetimetillcloggingbeinglonger.

Moreinsightisprovidedbythefluidtransportasdisplayedin Fig.5.Theeffectivewaterfluxesinindividualcellsatthepointof localprecipitation(closedsymbols)andattheoutlet(open sym-bols)are plottedagainst time. Thestreams aresplitinto water fluxintheCO2-richphase(bluecircles)andintheaqueousphase

2_Experimental_capillary _pressure _(p

C(SW)) andrelativepermeability satura-tionfunctions(kr(SW))wereapproximatedbythevanGenuchtenmodels(van Genuchten,1980)availableinTOUGH2.Weperformedthesimulationsusingthe followingparameterset:=0.48,Slr=0.079,Sls=1,1/P0=3.5×10−4andPmax=109 forpC(SW)and=0.7,Slr=0.15, Sls=1andSgr=10−3)forkr(SW). Furthermore, K=500mD,Pin=100bar,T=45◦C(isothermal)andXNaCl=0.24(initialbrinesalinity inwt%)wereused.

Fig. 4. TOUGH2 simulations of core-flood experiments. Upper panel: CO2 -saturationprofilesduringCO2injectionintoabrine-saturatedrocksampleatarate of3×10−5kg/sandatdifferenttimesteps(linesandfilledsquares).Alsoshown arethefinalstatesofsimulationsathigherinjectionratesasindicatedinthe fig-ure(opensymbols).Themiddleandlowerpanelsshowtherespectivebrineand solid-saltsaturationprofiles.

(blacksquares).Theredlinemarksthetotalwaterfluxattheoutlet. ThewaterfluxintheCO2-richphaseisdeterminedbythewater content andthe flow rate.It stays constant over a longperiod oftime,accordingtothesaturationlimit.Thewatertransportin theaqueousphaseisverystrongduringthefirstfewsecondsbut

Fig.5.H2Ofluxesinselectedgridblocks(gb)asafunctionoftime.Squaresrepresent theH2OfluxintheaqueousphaseandcirclesrepresenttheH2OfluxintheCO2-rich phaseduetoevaporationandsubsequentadvection.Notethattheaqueousphase changestheflowdirectionasindicatedbythearrows.(Forinterpretationofthe referencestocolorintext,thereaderisreferredtothewebversionofthearticle.)

(6)

ARTICLE IN PRESS

G Model

declinesrapidlyaccordingtothemobilityratioofthefluidphases. Theaqueousphasebecomesimmobileandthewatertransportis determinedbyevaporationandsubsequentadvectionintheCO2 -richphase.Afteracertaintime,capillarypressuredominatesthe aqueous-phasetransport,leadingtoacapillary-drivenbackflow intothezoneofevaporation–countercurrenttotheinjectedCO2. Backflow,however,allowslocalaccumulationofsaltbeyondthe brine’soriginallocalsaltcontent.Thepointatwhichthemajorpart ofthesaltprecipitatesisthepointatwhichthe‘evaporationfront’ nolongermoves.Here,thenegativewaterfluxinthebrinephase compensatesforthepositivewaterfluxintheCO2-richphase(see Fig.5)andwecanformulateaconditionforlocalprecipitationto occurif:

qSC·SC·XH2O,SC=qaq·aq·XH2O,aq, (3)

whereqSCandqaqarethevolumetricﬂuxesoftheCO2-richphase andtheaqueousphase,SCandaq arethephasedensities,and XH2O,SCandXH2O,aqarethemassfractionsofwaterinbothphases,

respectively.

Thesimulationsdemonstrateandexplaintheeffectobserved in the experiment. However, there are substantial differences betweenexperimentsandsimulationsrelatedtosaturation gradi-ents.Inthesimulation,thegradientisdominatedbythecapillary end-effectandisthereforeageometricalproperty.Inthe experi-ment,however,theendeffectissuppressed(asobservedintheCT profiles)bycapillarybackflow,i.e.capillaryredistribution;in con-trasttothesimulation,waterevaporatesovertheentirecorelength and,hence,theinducedbackflowequalizesthebrinesaturation, suppressingtheendeffect.Thisisaresultofevaporationkinetics, whichisnotaccountedforinthesimulationswithTOUGH2(orin anyothersimulatorwhichdoesnottakeevaporationkineticsinto account).Hencethesaturationgradientthatleadstolocal precip-itationisnotquantitativelymodeled.Second,thereisonlyafinite brinevolumeintheexperiment,duetothefinitesamplesize.This leadspotentiallytoanunderestimationofsaltsaturationasaresult ofdepletioneffectsinexperimentswithafinitevolume,whichcan beshownbysimulationswithafinitesamplevolumes(datanot shownhere).

Thecriticalflowrate,abovewhichahomogeneousprecipitation patternisobservedintheexperiment(i.e.atwhichtheadvective fluxpreventscapillary-drivenbackflow),isequivalenttotheratein thesimulationsatwhichtheprecipitationpeakmovesawayform theinlettothenextpointwithahighersaturationgradientwerethe downstreamfluxofwaterinCO2isagaincompensatedby counter-currentflowoftheaqueousphase.Suchashiftisnotobservedfor theflatsaturationprofilesintheexperiments,sincetheconstant saturationgradientcannotcounteracttheadvectiveforceatflow ratesabovethecritical.Aflatsaturationprofileisexpectedinthe field.However,inradialflowgeometry,wheretheadvectiveforce decreaseswithdistance,theconditionoflocalprecipitationcanbe reachedatacertaindistanceevenwithaflowrateabovethecritical closetotheinjectionpoint.

5. Thezoneofattraction:kineticmodel

InthesimulationspresentedsofartheproducedCO2wasfully water-saturatedasresultoftheequilibriumapproachofthe evap-orationmodel.However, fromtheexperimentsin Section3 we determined theaverage water saturation of the produced CO2 streamtobeabout35%ofthesolubilitylimitfortherespective experimentalconditions.Thereasonforthepartialsaturationis theﬁnitecontactareabetweentheinjectedCO2streamandthe brine in therock and the ﬁnite evaporationrate,which is not takenintoaccount(overestimated)intheequilibriummodel.On thebasisoftheaveragewatercontentoftheproducedCO2,we

Fig.6.Zoneofpotentialattraction.Upperpanel:WatersaturationintheCO2-rich phaseasfunctionofdistancefromtheinjectionpoint.Theprofilesfordifferent injectionratesareshownsimulatedinalinearflowgeometry.Lowerpanel:The zoneofattractionasafunctionofinjectionrate.Thefilledsymbolsareobtained fromexperiments.

estimatedthezoneinwhichevaporationtakesplacetobeinthe orderof0.3–1minlineargeometry–inanycaselongerthanthe experimentallengthscale.Inthiszone,thewatersaturation gradi-ent(dSW/dx)ismodiﬁedfromthegradientexpectedfromviscous displacementonlyandcapillary-drivenbackﬂowcanoccur.The evaporationzoneisthereforethezoneoverwhichcountercurrent solutetransportcanpotentiallyoccur–wethereforecallitzoneof attraction.

Inthis sectionwepresentan estimationofthezone dimen-sion in linear geometry by numerical simulations. We use a kinetic approach as reported byRoels et al. (2014) and model thezoneofattractionfor differentinjectionratesbetween0.01 and 100ml/min/cm2 ₍₂_×₁₀−6 _and _0.1_m/s). _We _use _rock-ﬂuid parameters–relativepermeabilityandcapillarypressure satura-tionfunctions–obtainedforCO2-brinedisplacementinthesame rocktype(Bereasandstone:Ottetal.,2011a;Bergetal.,2013),for thesameﬂuidcombinationandthermodynamiccondition.

We model evaporation and ﬂow only and we do not take thefeedbackofprecipitationonpermeabilityandcapillarityinto account.Wedeterminetheoverallevaporationrate–A×revap,with AbeingtheCO2-brinecontactareaandrevaptheevaporationrate –bymatchingtheaveragewaterconcentrationintheproduced CO2streamtotheexperimentallyobtainedvalueattherespective distance.Subsequently,simulationswereperformedwith experi-mentalinputdataonasemi-inﬁnitesimulationdomain.Fromthe simulationdatawedeterminedthedistance,levap,overwhich evap-orationtakesplace–i.e.thedistancethatisneededtoreachwater saturationlimitinCO2undertherespectiveconditions.

Fig.6showsthewatersaturationproﬁlesintheCO2streamfor differentinjectionrates.Eachsimulationshowsthetypicalincrease ofthewatersaturationintheCO2phase(XH2O,SC)withdistance

(7)

Fig.7. ␮CTdatarecordedwithpseudopore-scaleresolutionduringCO2ﬂooding.Toprow:cross-sectionaldifferenceimagesshowingthesolidsaltsaturationintheregion oflocalprecipitationbetween0and0.5cmfromtheinlet(imagecatmaximumsaltsaturation).Lowerimage:3Drepresentationofthesaltdistributionat0.25cmfromthe inlet(orange-red).TheinitialCO2percolationpattern(after0.4h)isaddedattherightfrontcornerofthe3Dvolume.

Thedatashowanincreasingextendofthezoneofattractionwith increasinginjectionrate.Thelowerpanelshowsthederivedlevap asafunctionofCO2injectionrate.Theexperimentallyestimated dataareincludedin theplot.Forthedeﬁnedrange ofinjection rates,evaporationzonesintheorderoflevap=3×10−3_to_above₁₀_m werefoundwithalineardependenceontheinjectionrate.Forthe conditionsatwhichtheexperimentswerecarriedout,thezoneof attractionhasbeensimulatedtobe0.4m.

The presented kineticsimulations give valuable information aboutthesizeofthezoneaffectedbysalt precipitationandthe distancesoverwhichsolutescanbetransportedcountercurrentin directionoftheinjector.Thisservedthereservoirengineerasinput fore.g.griddingofreservoirmodelsaroundgasinjectorsand pro-ducers.Thesaltvolumethatpotentiallycanbetransportedtothe pointofdry-outcanbeestimatedfromtheimmobilebrinefraction intherespectivelyaffectedvolume.

6. Effectsonpermeability:K()relationship

Up to this point we have discussed the longitudinal solute transport and theresulting precipitationproﬁles. However,the importantquestionisrelatedtotheeffectofprecipitation–i.e. porosityreduction–onpermeabilityandeventuallyoninjectivity. Whileporosityreductionismeasuredby␮CT,changesof perme-abilityarereﬂectedinthepressuredropP.

TheupperpanelofFig.8showsthepressuredropoverthecore lengthduringflooding(blackline).Pessentiallydecreases dur-ingtheexperiment.Theonsetofcapillarybackflowisvisibleasa smalldip,withasubsequentPincreaseduetosubstantiallocal precipitation.Butdespitethisporosityreduction,theeffective per-meabilityoftheCO2-richphaseincreasedduringtheexperiment, whichmeansthatrelativepermeabilityeffectscompensateforthe absolutepermeability (K)reduction. A Pcurve ofa compara-bleexperimentthatshowedahomogeneousprecipitationpattern withaboutthesameaverageporosityreductionisshowninthe sameplot(redline).Theexperimentwasterminatedafterabout 7handshowsasmallerfinalreductionofKasintheheterogeneous case.Fromthefinalporosityprofilesofbothexperimentsasshown inFig.3andtherespectivepressuredrop,wedetermineaK()

relationshipforthedrystatebyusingthecorrelationproposedby VermaandPruess(1988):

K−KC K0−KC =

₋_C 1−C

, (4)

withKCbeingthelowestvaluetowhichKcanbereducedby pre-cipitation.WhileKC isarobustoutcome,theexponent andthe criticalporosityCarenotuniquelydeﬁnedbythedatasetalone. However,thetwoextremes,C=0andC=0.8,givesimilarresults

Fig.8.Top:PressuredropPoverthecoreduringCO2injection.Twocasesare shown:(1)thelocalprecipitationcaseasdiscussedinthetext(blackline),and (2)Precordedduringhomogeneousprecipitationforcomparison(redline).The opensymbolsrepresentthedevelopmentofthelocalprecipitate(saltsaturationin arbitraryunits)inFig.3.Bottom:K()relationshipsdescribingtheexperimental observations.

(8)

ARTICLE IN PRESS

G Model

Fig.9. Schematicviewonthedry-outandprecipitationprocess.(a)Retractionofthebrinephase(menisci)duetoevaporationleadingtoanincreaseofCO2relative permeability.(b)Theprecipitateisessentiallylocatedinthevolumeformerlyoccupiedbythebrinephase.

(withtheexponentﬁttedtothedata)asshowninthelowerpanel ofFig.8 suchthateitherresultcanbeusedforapractical pur-poseintherelevantporosityrange=/0=1to0.8.Aswewill arguebelow,themaximumsaltsaturationisequaltotheresidual brinesaturation(Ssalt,max=SW,res)andthereforeC=1−SW,res=0.8. Withthisweareabletodeterminetheremainingﬁtting parame-ter ,resultingintheparameterset:KC/K0=3.5×10−3,C=0.8and =10.1.

KCislikelytobearesultoftheobservedprecipitationpatternas showninFig.7a–e.Theimagesshows␮CTcrosssections(toprow) anda3DimageofthedataalreadypresentedinFig.3andwere recordedwith24␮m voxelsize (pseudopore-scale resolution), whichisacompromisebetweenspatialresolutionandaproper ﬁeldofviewwithanimprovedtimeresolution(shortertotal scan-ningtime).Theimagesrangefromclosetotheinlet(a)toabout 0.5cmintothesample(e)andshowthesolid-saltcontributionof thedrysample.Conspicuously,thesaltprecipitatesinspotson dif-ferentlengthscales,rangingfromamm-scaletoasub-mmoreven amicrometer-scale.A3Drepresentationofthesolidsalt distribu-tionisshowninthelowerpartofFig.7inorange,whiletheearly CO2 percolationpatternisgiven intherightfrontcornerofthe 3Dgraphinblue.Bothpatternsoccupycomplementaryspaceand overlapbylessthan5%ofthecommoncrosssection.Thisindicates thatsaltprecipitatesinthevicinityoftheinitiallyCO2occupied volume,leavingtheinitialCO2-pathwaysessentiallyopen.

Thedatapresentedsofartogetherwithacoupleofsimple argu-mentsallowustoconstructamindmodel–illustratedinFig.9– thatsupportsandsummarizestheﬁndingsofthiswork:

1.Viscousbrinedisplacementandwaterevaporationactin par-allel,butwerefoundtobedominantondifferenttimescales. Whileresidualbrinesaturationisusuallyreachedafter injec-tionofafewporevolumes,completedry-outwasreachedafter hundredsofporevolumesofinjectedCO2.Thereisessentially noviscousbrinedisplacementduringdry-out.

2.Thestructureoftheresiduallytrappedbrinephaseisdetermined bycapillaryforces.

3.Afterreachingthesolubilitylimit,saltprecipitatesinthebrine phase,andhenceinthevolumeoccupiedbytheresidualbrine, SW,res.

4.Thebrine-saturatedvolumeretractsduringevaporation,leaving theprecipitatedsaltbehind.

5.Thereisnotransportmechanismofliquidwatervaporacross theCO2–brineinterfaceandhencenotransportmechanismof saltintotheintotheCO2ﬂowchannels–theﬂowchannelsserve forwatervaportransport,butstayopensincesaltprecipitates intheresidualbrinephase.

brine

CO2

q

SC

>

cr

q

SC

~

q

cr

q

SC

<

q

cr

uniform prec. local prec. prec. at inj. point

capillary-driven

back flow

evaporation

flow

channels

Fig.10. Toprow:schematicviewofthedifferentflowregimesthatdeterminethe precipitationpatternandarecharacterizedbytheflowratewithrespecttoacritical flowrateqcr.Bottomrow:schematiccross-sectionalcapillary-drivenbrinetransport asdescribedinthetext.

6.Becauseofthewellseparatedtimescalesofviscous displace-mentandevaporation,themaximumporevolumethatcanbe ﬁlledbyprecipitateisthevolumecorrespondingtotheresidual brinesaturation,SW,res.We conclude:Ssalt,max=SW,res.InBerea sandstonewetypicallyﬁndSW,res≈0.2,whichcorrespondsto themaximumobservedsaltaccumulationinthepresentstudy.

7. Summaryandconclusions

Insummary,weinvestigatedconsequencesofdry-zone forma-tioninthevicinityofgasinjectionandproductionwells.Incore flood experimentswe injected dry SCCO2 into brine-saturated sandstone samples and reached complete dry-out. During the floodswemeasuredeffectivepermeabilityandfluidandsalt satu-rations.

Wefoundthattimescalesofviscousdisplacementanddrying arepracticallywellseparates.Twomechanismsturnedouttobeof importance:(1)macroscopicsolutetransportduetocapillary pres-suregradients,whichdeterminesthemacroscopicsaltdistribution, and(2)thepore-scalearrangementoffluidphasesduringviscous displacementandthedryingprocess.Suchlocaleffectsdetermine effectivepermeability,theporosity–permeabilityrelationshipand probablythemaximumpossiblesaturationofsaltintheporespace. Theprincipalmechanismofsolutetransportandthecondition underwhichlocalprecipitationoccurshavebeeninvestigatedby experimentsandnumericalsimulations.Theresultsare schemati-callydisplayedinFig.10.Weidentifiedtheoriginofacriticalflow rateabove whichsalt precipitateshomogeneously ona macro-scopicscale.Below thecritical flowrate,capillary-induced back

(9)

flowoftheremainingbrinephasetransportssolutesinthe direc-tionoftheinjectionpoint,wheresalteventuallyprecipitates.We furtheridentifiedthelengthscaleoverwhichsolutescan poten-tiallybetransported asafunctionofinjectionrate.Thezoneof attractionhasbeenfoundtobe0.4mforthedescribed experimen-talsituationandcanreachseveralmetersdependingontheflow rate.Duetocountercurrentflow,theamountofprecipitateper unitvolumecanexceedthevolumeofsaltoriginallydissolvedin thebrineintherespectivevolume.

From the presented experiments we calculated the permeability–porosity relationship, K(), describing the effect ofprecipitationinsandstoneundertherespectiveﬂowconditions. Despitetheobservedstrongreductionsofabsolutepermeability, theeffectiveCO2permeabilityincreasedduringtheexperimentin allcases.Thisrelativelymildimpactofprecipitationonthesample permeabilitycanbeattributedtotheobservedlocalprecipitation pattern;saltprecipitatesinthebrinephaseandhenceinthe vicin-ityofCO2-conductingchannels,leavingthesechannelsessentially open.

Theresultsofthestudyareofdirectrelevancefortherisk assess-mentofinjectionoperations,andareofpracticalusedforinjectivity modeling.Forthepredictionofmacroscopicsalttransportandthe resultingporosityreductions,injectionratesandeffective evapo-rationratesneedtobetakenintoaccount.Thedimensionofthe affectedzone hasbeen estimatedto bein theorder of several centimeterstometers,whichisvaluableinputforanadequate grid-dingaroundgasinjectorsandproducersinreservoirmodeling.For similarrocktypes,therespectivepermeabilityreduction canbe modeledbytheobtainedK()relationship.

Saltprecipitationinsimilarrocktypeswillleadtoan absolute-permeabilityreduction,buttoanimprovingeffectivepermeability duringtheinjectionprocess.Forthatreason,weexpectarelatively mildimpactofsaltprecipitationoninjectivitycomparedtothecase ofmulti-modalcarbonatesasinvestigatedinacompanionpaperby Ottetal.(2014).

Acknowledgments

The authorsthank Steffen Bergand JeroenSnippe for fruit-fuldiscussionsand forreviewingthemanuscript.Fons Marcelis isacknowledgedforsamplecharacterizationandpreparationand PacelliZithaforcontinuoussupportanddiscussions.

References

Andre,L.,Peyssona,Y.,Azaroual,M.,2014.WellinjectivityduringCO2storage

oper-ationsindeepsalineaquifers:Part2.Numericalsimulationsofdrying,salt depositmechanismsandroleofcapillaryforces.Int.J.Greenh.GasControl22, 301–312.

Bacci, G., Korre, A., Durucan, S., 2011. Experimental investigation into salt precipitation during CO2 injection in saline aquifers. Energy Procedia 4,

4450–4456.

Bachu,S.,Gunter,W.D.,2004.Overviewofacid–gasinjectionoperationsinWestern Canada.In:Proceedingsofthe7thInternationalConferenceonGreenhouseGas ControlTechnologies,Vancouver,Canada,5–9September,2004,vol.1.

Berg,S.,Oedai,S.,Ott,H.,2013.Displacementandmasstransferbetweensaturated andunsaturatedCO2–brinesystemsinsandstone.Int.J.Greenh.GasControl12,

478–492.

Ehrenberg,S.N.,Nadeau,P.H.,2005.Sandstonevs.carbonatepetroleumreservoirs:a globalperspectiveonporosity-depthandporosity–permeabilityrelationships. AAPGBull.89(4),435–445.

Giorgis,T.,Carpita,M.,Battistelli,A.,2007.2Dmodelingofsaltprecipitationduring theinjectionofdryCO2inadepletedgasreservoir.EnergyConvers.Manag.48,

1816–1826.

Hubbell,J.H.,1969.PhotonCrossSections,AttenuationCoefﬁcients,andEnergy AbsorptionCoefﬁcientsFrom10keVto100GeV.Tech.Rep.NSRDS-NBS29.U.S. DepartmentofCommerce,NationalBureauofStandards.

IPCC,2005.IPCCSpecialReportonCarbonDioxideCaptureandStorage.Cambridge UniversityPress,UK.

Kleinitz,W.,Koehler,M.,Dietzsch,G.,2001.Theprecipitationofsaltingasproducing wells.SocietyofPetroleumEngineersSPE68953,1–7.

Lide,D.R.,2003.HandbookofChemistryandPhysics.CRCPress.

Ott,H.,Berg,S.,Oedai,S.,2011a.DisplacementandmasstransferofCO2/brinein

sandstone.In:SocietyofCoreAnalysisConferencePaperSCA2011-05.

Ott,H.,deKloe,K.,Marcelis,F.,Makurat,A.,2011b.InjectionofsupercriticalCO2in

brinesaturatedsandstone:patternformationduringsaltprecipitation.Energy Procedia4,4425–4432.

Ott,H.,deKloe,K.,Taberner,C.,Marcelis,F.,Wang,Y.,Makurat,A.,2010.Rock/ﬂuid interactionbyinjectionofsupercriticalCO2/H2S:investigationofdry-zone

for-mationneartheinjectionwell.In:SocietyofCoreAnalysisConferencePaper SCA2010-20.

Ott,H.,deKloe,K.,vanBakel,M.,Vos,F.,vanPelt,A.,Legerstee,P.,Bauer,A.,Eide, K.,vanderLinden,A.,Berg,S.A.M.,2012.Core-ﬂoodexperimentfortransportof reactiveﬂuidsinrocks.Rev.Sci.Instrum.83(084501-1-084501-16).

Ott,H.,Snippe,J.,deKloe,K.,Husain,H.,Abri,A.,Makurat,A.,2014.Saltprecipitation duetosupercriticalgasinjection:II.Singlevs.multiporosityrocks.Int.J.Greenh. GasControl(submittedforpublication).

Pape,H.,Clauser,C.,Ifﬂand,J.,1999.Permeabilitypredictionbasedonfractal pore-spacegeometry.Geophysics64(5),1447–1460.

Perrin,J.-C.,Benson,S.,2010.Anexperimentalstudyontheinﬂuenceofsub-core scaleheterogeneitiesonCO2distributioninreservoirrocks.Trans.PorousMedia

82,93–109.

Peyssona,Y.,Andre,L.,Azaroual,M.,2014.WellinjectivityduringCO2storage

opera-tionsindeepsalineaquifers:Part1.Experimentalinvestigationofdryingeffects, saltprecipitationandcapillaryforces.Int.J.Greenh.GasControl22,291–300.

Pruess,K.,García,J.,2002.MultiphaseﬂowdynamicsduringCO2injectionintosaline

aquifers.Environ.Geol.42,282–295.

Pruess,K.,Müller,N.,2009. Formationdry-outfrom CO2 injectioninto saline

aquifers:1.Effectsofsolidsprecipitationandtheirmitigation.WaterResour. Res.45,W03402.

Pruess,K.,Spycher,N.,2007.ECO2N–aﬂuidpropertymodulefortheTOUGH2 codeforstudiesofCO2storageinsalineaquifers.EnergyConvers.Manag.48,

1761–1767.

Roels,S.M.,Ott,H.,Zitha,P.L.J.,2014.␮-CTanalysisandnumericalsimulationof dryingeffectsofCO2injectionintobrine-saturatedporousmedia.Int.J.Greenh.

GasControl27,146–154.

Schutjens,P.M.T.M.,Hanssen,T.H.,Hettema,M.H.H.,Merour,J.,deBree,P., Core-mans,J.W.A.G.H.,2004.Compaction-inducedporosity/permeabilityreduction insandstonereservoirs:dataandmodelforelasticity-dominateddeformation. SPEReserv.Eval.Eng.7(3),202–216.

Span,R.,Wagner,W.,1996.Anewequationofstateforcarbondioxidecovering theﬂuidregionfromthetriple-pointtemperatureto1100katpressuresupto 800mPa.J.Phys.Chem.Ref.Data25(6),1509–1596.

Spycher,N.,Pruess,K.,2005.CO2–H2Omixturesinthegeologicalsequestrationof

CO2:II.Partitioninginchloridebrinesat12–100◦Candupto600bar.Geochim.

Cosmochim.Acta69(13),3309–3320.

vanGenuchten,M.T.,1980.Aclosed-formequationforpredictingthehydraulic conductivityofunsaturatedsoils.SoilSci.Soc.Am.J.44,892–898.

Verma,A.,Pruess,K.,1988.Thermohydrologicalconditionsandsilicaredistribution nearhigh-levelnuclearwastesemplacedinsaturatedgeologicalformations.J. Geophys.Res.93,1159–1173.

Vinegar,H.J.,Wellington,S.L.,1987.Tomographicimagingofthree-phaseﬂow experiments.Rev.Sci.Instrum.58(1),96–107.

Wellington,S.I.,Vinegar,H.J.,1987.J.Petrol.Technol.SPE16983,885–898.

Wyble,D.O.,1958.Permeabilitypredictionbasedonfractalpore-spacegeometry. Trans.Soc.Pet.Eng.AIME213,430–432.

Xu,T.,Apps,J.A.,Pruess,K.,2004a.NumericalsimulationofCO2disposalbymineral

trappingindeepaquifers.Appl.Geochem.19,917–936.

Xu,T.,Ontoy,Y.,Molling,P.N.S.,Parini,M.,Pruess,K.,2004b.Reactivetransport modeling ofinjectionwellscalingandacidizing atTiwiﬁeld,Philippines. Geothermics33,477–491.

Zuluaga,E.,Monsalve,J.C.,2001.Watervaporizationingasreservoirs.Soc.Petrol. Eng.SPE84829,1–4.

Zuluaga,E.,Mu ˜noz,N.I.,Obando,G.A.,2001.Anexperimentalstudytoevaluatewater vaporisationandformationdamagecausedbydrygasﬂowthroughporous media.Soc.Petrol.Eng.SPE68335,1–7.