Application of a novel approach to modelling the supercritical extraction kinetics of oil from two sets of chia seeds

Application

of

a

novel

approach

to

modelling

the

supercritical

extraction

kinetics

of

oil

from

two

sets

of

chia

seeds

David

Villanueva-Bermejo

a,

*

,

Tiziana

Fornari

a

,

Maria

V.

Calvo

a

,

Javier

Fontecha

a

,

Jose

A.P.

Coelho

b,c

,

Rui

M.

Filipe

b,d

,

Roumiana

P.

Stateva

e

a_{InstitutodeInvestigaciónenCienciasdelaAlimentaciónCIAL(CSIC-UAM),28049Madrid,Spain} b

InstitutoSuperiordeEngenhariadeLisboa,InstitutoPolitécnicodeLisboa,1959-007Lisboa,Portugal

c

CentrodeQuímicaEstrutural,InstitutoSuperiorTécnico,UniversidadedeLisboa,1049-001Lisboa,Portugal

d

CERENA,CentrodeRecursosNaturaiseAmbiente,InstitutoSuperiorTécnico,UniversidadedeLisboa,1049-001Lisboa,Portugal

e

InstituteofChemicalEngineering,BulgarianAcademyofSciences,1113Sofia,Bulgaria

ARTICLE INFO Articlehistory: Received7January2019

Receivedinrevisedform11October2019 Accepted25October2019

Availableonline12November2019 Keywords:

Chiaseedoil

SupercriticalCO2extraction

Polyunsaturatedfattyacids(PUFA) α-Linolenicacid(ALA)

Extractionkinetics Mathematicalmodelling

ABSTRACT

Thekineticsofthesupercriticalfluidextractionofedibleanddiscardedchiaseedswasstudiedand

modelledforthefirsttime.Thetotaloilwasremovedat45MPaand60
Cafter240min.Theextraction

kineticswassimulatedusingadynamicmodelingPROMSModelBuilderenvironmentandthekinetic

parametersestimated.Trioleinwaschosenasamodelcompoundofthechiaoil.Theagreementbetween

theexperimentalyieldsand thosecalculatedbythemodelwasgood withdeviationsintherange

(1.2–6.6)%,exceptat25MPaand60
_{C(AARD=9.5%).}

©2019TheKoreanSocietyofIndustrialandEngineeringChemistry.PublishedbyElsevierB.V.Allrights

reserved.

Introduction

Atpresent,thereisaconstantsearchofpromisingandinexpensive vegetalmaterialsasasourceofpolyunsaturatedfattyacids(PUFAs). Chia (Salvia hispanica L.), a plant indigenous to Guatemala and Mexico,whichbelongstotheLamiaceaefamily,isattractingagreat attention,especiallytheseeds.Chiaseedsarebeingstudiedasa sourceoffoodingredients,suchasproteinsanddietary_fiber[1_–3], buttheymainlystandoutfortheirhighoilcontent(20–35mass%). Chiaseedoilcontains significantamounts ofPUFAs,mainlythe omega-3α-linolenicacid(ALA)(60–65%oftotalfattyacids)[4], whichroleinthepreventionofcardiovascular,nervoussystemand inflammatorydiseaseshasbeenthoroughlydescribed[5,6].

Thus far, different solvents (ethyl acetate, acetone, propane, petroleumetherandhexane)andextractiontechniques(coldand hotpressing,Soxhlet,Ultrasound-AssistedExtractionandPressurized LiquidExtraction)havebeenappliedtoobtainchiaseedoil[7–11].A viableandeco-friendlyalternativetotheuseoforganictoxicsolvents istheextractionwithsupercriticalCO2(scCO2),whichisnon-toxic,

non-flammable,non-mutagenicandcarcinogenicandisabundant

andinexpensive.Moreover,duetothepossibilityofworkingatlow temperaturesandintheabsenceofoxygen,supercriticalextraction withscCO2(SCE)preventsorminimizesthedegradationofbioactive

compounds andallowsobtainingsolvent-free products[12].New developmentsregardingtheapplicationofthisadvancedtechniqueto obtainoilswerereportedrecentlyintheliterature.Forexample,Wei etal.[13]employedultrasound-assistedsupercriticalcarbondioxide extractionforremovingoleanolicacidandursolicacidfromHedyotis corymbose.Theexperimentalsolubilitydata,calledbytheauthors fictitious, were read from the initialslope of the curve of the

extraction yield versus the amount of scCO2 used, and were

modelledapplyingseveralsemi-empiricaldensity-basedmodels. Moonetal.[14]studiedthescCO2extraction,withandwithout

co-solvent(ethanol)oftheessentialoilfromAsiasarumheterotropoides

and the results obtained were compared with conventional

extraction.Inanotherstudy[15],turmeric(CurcumalongaL.)was

extracted withscCO2 and turmerones wereconcentrated using

semi-preparativesupercriticalchromatography.Supercriticalfluid extractionwithaco-solventwasalsoappliedtoextractoilfromrice branwiththeaimtopromotethevalorizationofthisabundant feedstock[16].ConcerningSCEofchiaseedoil,asfarasweare aware,onlyafewstudieshavebeencarriedouttillpresent[17–20]. Inthose,theconcentrationofALAachievedintheextractedoilwas approximately (60_–65) %,hence SCE processallowed obtaining

* Correspondingauthor.

E-mailaddress:[email protected](D.Villanueva-Bermejo).

https://doi.org/10.1016/j.jiec.2019.10.029

1226-086X/©2019TheKoreanSocietyofIndustrialandEngineeringChemistry.PublishedbyElsevierB.V.Allrightsreserved.

ContentslistsavailableatScienceDirect

Journal

of

Industrial

and

Engineering

Chemistry

healthy and high-quality oils regardless of the operational parametersapplied.

Itisknownthatkineticdataareessentialfortherealizationofa feasibleindustrialprocess. However,despitetheaforementioned andthepossibilitiestorealizea viableindustrialprocess,kineticdata arenotonlyscarceandsuperficial,butalso,asfarasweareaware, therearenoattemptsrelatedtothemodellingofSCEofoilfromchia seedsreportedintheliteraturetillpresent.Accordingly,twowere theobjectivesofthiswork:i)toprovidenewdataonthescCO2

extraction of chiaoil and ii) to applya novel approach tothe extractionkineticsmodelling,advocatedoriginallybySovováand Stateva[21].Thisapproachre_flectstheinteractionbetweenkinetics andphaseequilibria(solubility)andappliesareliableandversatile modellingframeworktoestimatethesolubilityof theoilinthe supercriticalfluid,asdiscussedindetailinSection“Supercritical fluidextractionofchiaoil—modellingframework”.

Toachievetheobjectivesofourwork,ediblechiaseeds(ECS), anddiscardedseeds(DCS)werestudied.Particularlyforthelatter, thepresentinvestigationcanbeofconsiderableinterestasitwill provideinformationonhowtointensifyfurthertheirvalorization and industrial applications. So far, the studies reported in the literaturehavebeencarriedoutbyusingECS.Chiaseedsemployed inthepresentstudyweresubjectedtoaselectionprocessinwhich theseedswereclassifiedin differentqualitiesmainlybased on theirweight,intactness,color,visualaspectandsize.Thus,DCS

consist of damaged, partially broken and/or smaller-size and

lower-weightseedsthatareusuallydiscardedduringpost-harvest handlingandfinallyintendedtoanimalfeeding.Theirprice,asa consequenceofthemarketsurplusofthisproduct,isconsiderably lowerthanthatofECS.Eventhoughtheyareanunderutilizedraw

material, DCS possess a noteworthy amount of oil and may

constitute a viable alternative source for obtaining highly

polyunsaturatedchiaseedoilstobeusedin humanhealthcare formulations.

Materialsandmethods

Samplepreparation

The two different sets of chia seeds (Salvia hispanica L.)

originating fromMexico were purchased from a local supplier

(Primaria).Theoilcontentforeachsetofseeds(28.3%and19.9% massforECSandDCS,respectively)wasdeterminedby pressur-izedliquidextractionbyusinga2:1chloroform:methanolmixture at60
C and10minofextractiontime [22].Theseresultsarein agreementwiththevaluesprovidedbythesupplierforeachset (25.2%and20.6%,respectively).Inwhatfollowswewillusetheoil contentvaluesobtainedbyusforeachsetofseeds.

Theseedsweregroundinaknifemillcooledbyliquidnitrogen. Groundseedsweresievedusingmeshsizesof0.250and0.500mm Ø (CISACedaceria Industrial S.L. Barcelona, Spain). An average particlediameter(dp)of0.370mmwasobtained,usingEq.(1):

dp¼ Mt Pj i¼1 mi dpi ð1Þ

whereMtisthetotalmassofmilledseeds,mi —themassofthe particleskeptbelowmeshsizedpi andj—thenumberofmesh sizes.

Samplesobtainedwerestoredat20
_{Cuntiltheiruse.} Chemicals

Dichloromethane,hexane,methanol,acetonitrile,

dimethylfor-mamide (HPLC grade) and sulfuric acid (98% purity) were

purchased from Labscan (Dublin, Ireland). Sodium carbonate,

seasandandsodiumsulfateanhydrousweresuppliedbyPanreac (Barcelona,Spain).Sodiummethoxide(95%purity)wassupplied by Sigma–Aldrich (St. Louis, MO, USA). Butterfat BCR-164 (EU

Commissions; Brussels, Belgium) was supplied by FedelcoInc.

(Madrid, Spain). CO2 (99.99% purity) was supplied byCarburos

Metálicos(Madrid,Spain). SupercriticalCO2extraction

The SCEswereperformed ina pilot-plantsupercritical fluid extractor(modelSF2000;TharTechnology,Pittsburgh,PA,USA), equippedwitha273cm3_{cylinderextractioncell(18.8cmlongand}

4.3cminternal diameter)and twoseparators(0.5Lcapacity).A thoroughdescriptionoftheequipmentcanbefoundin Villanueva-Bermejoetal.[23].

TheSCEsfromDCSwerecarriedoutatp=(25and45)MPaand T=(40and60)o_C.TheCO

2flowrate andextractiontimeswere,

respectively,40gmin1and240minforalltheexperimentswith thissetofseeds.TheextractionsofECSwereperformedat45MPa and40
C.SeveralCO2flowrateswerestudied,namely27,40and

54gmin1(CO2-to-seedratioof50,74and100,respectively).For

allruns,theseedsmassusedwas130g,withanapparentdensity valueof0.6060.002gcm3.DuringtheexperimentsthescCO2

wasrecirculated.Theentireextractswerecollectedfromthefirst separator(themassofoilobtainedfromthesecondseparatorwas negligible) by depressurization at 5MPa (system recirculation

pressure). Oil samples were dissolved in methylene chloride,

treated with  1g of sodium sulfate anhydrous and filtered

through0.45

m

mfilters.Finally,thesampleswerestoredat35
_C untilanalysis.

Fattyacidprofile

Thederivatizationoffattyacidsfromchiaoilswascarriedout followingthemethoddescribedbyCastro-Gómezetal.[24].The analysisoffattyacidmethylesters(FAMEs)wereperformedinan Agilentchromatograph6890N(AgilentTechnologiesInc.PaloAlto, USA)equippedwithanMSdetector(Agilent5973N)andusing CP-Sil88fused-silicacapillarycolumn(100m0.25mmID0.2

m

m.

Chrompack, Middelburg, The Netherlands). The temperature

program started at 100
C for 1min and then the temperature increasedby7
Cmin1upto170
C, followedbyanisothermal periodof55min.Finallytemperatureincreasedby10
Cmin1up to230
Candwasheldfor33min.Theinjectortemperaturewas 250
Candheliumwasusedasthecarriergas.Theanalysiswas carriedoutinsplitmode(splitratio1:25)andtheinjectionvolume was 1

m

L. For the MS detector, the transfer line, source and

quadrupole temperatures were 250
C, 230
C and 150
C,

respectively. The mass spectrometer operated under electron

impactmode(70eV).Elutingcompoundswerescannedintotalion current(TIC)modeinthemassrangefrom40to500mz1.The identificationoftargetcompoundswascarriedoutbycomparing theirmassspectrawiththoseattheNationalInstituteofStandards

and Technology (NIST) library (Gaithersburg, MD, USA). The

response factors were calculated using anhydrous milk fat

(referencematerialBCR-164).Tritridecanoin(C13:0)wasusedas aninternalstandard.

Supercriticalfluidextractionofchiaoil—modellingframework Modeldescription

The model developed by Sovová and Stateva [21] for

multicomponent systems was used in this work. In brief, the

approachreflectstheinterplaybetweensolubilityandkinetics.For

estimatethesolubilityoftheoilinthesupercriticalfluid,andthe

resulting equations are incorporated into the dynamic model.

Dynamicsimulationofthesupercriticalextractionprocessisthen performed.Themodelconsidersthattheconcentrationsinsidethe extractorarehomogeneousinthefluidandsolidphases.Internal diffusionisneglectedbasedontheassumptionthattheextracts arelocatedatthesurfaceofthesolidparticles,andhenceeasily available.

Asthechiaseedsusedinourworkweregroundedtoanaverage diameterof0.370mm,itwasconsideredthattheresultinginternal diffusionpathforsuchsmallparticlesisshort.Consequently,the oiliseasilyavailableattheparticlesurface.

Themodelisdescribedbythefollowingsetofequations: dw  dt þ w  tr ¼ kfa0

e

ðwþwÞ ð2Þ dws  dt ¼q 0 tr kfa0

e

ðwþwÞ ð3Þ wþ¼Kwsþ wb s wb tþwbs ðwsatKwsÞ ð4Þ

withtheinitialconditions:

wð0Þ¼w0 ð5Þ

wsð0Þ¼ws;0 ð6Þ

Theyield,e(kgkg1solid),isdefinedby: e¼q0

Z t 0

edt ð7Þ

eð0Þ ¼0 _ð8Þ

where w is theoil concentrationin thefluid phase inside the extractor(kgkg1CO2),ws—theoilconcentrationinthesolidphase

(kgkg1solid),tandtraretheextractionandresidencetime(min),

respectively,q’—thespecificflowrate(kgCO2min1kg1solid),

w+_{—theoilconcentrationatsolid-fluidinterface(kgkg}-1_CO 2),

e

—voidfractioninthebed,kfa0(min1)—thevolumetricfluid

phasemasstransferresistance,K(kgplantkg1CO2)—thepartition

coefficient,wt(kgkg1CO2)—themonolayeradsorptionmaximum

content,wsat(kgkg1CO2)—thesolubilityofthefreeoilcompound,

andbisacoefficientthatshouldbehigherthanone.

The model was deployed in gPROMS ModelBuilder [25], an

equation-oriented modelling environment for dynamic (and

steady-state)simulationthatincludesoptimizationandparameter estimationcapabilities.Itshouldbenotedthattheuseofthiskind

of equation-oriented modelling and optimization software for

supercriticalCO2extractionhasnotbeenthatmuchreportedinthe

literatureuntilnow.

Some of the coefficients in the model are unknown and

parameterestimation wasperformedtoestimatethesemissing

values. The simulated yield pro_files were compared with the

experimentaldata,andanobjectivefunctionwasusedtominimize theerroroftheadjustment,andobtaintheparametervaluesthat result in the best fit. Obtaining the solution of the resultant

nonlinear dynamic model may be challenging and may cause

numerical convergence problems. Also, locating the global

optimum is not guaranteed. A shortage in experimental data

may also compromise the reliability of the results and the

confidenceintervalassociatedwiththesolutions.

In this work, gPROMS ModelBuilder parameter estimation

capabilitieswereusedtoobtaintheunknownparameters.gPROMS

parameter estimation uses a maximum likelihood problem to

obtainthemissingparameters.Theinterestedreadercanfindmore detailsongPROMSdocumentation[25].

To beabletorelatethefittingaccuracyachieved,a standard

measure of deviation was used, the absolute average relative

deviation,AARD,definedby:

AARD¼100_N X N j¼1 eexp_i eest i   eexp_i ð9Þ

whereNisthetotalnumberofexperimentallymeasuredpoints, eexp_i and eest

i are the i-th experimental and estimated point,

respectively.

RepresentationoftheoilandcorrelationofitssolubilityinscCO2

Thechiaseedoil,asanyothervegetableoil,isaverycomplex mixtureofmanycompounds,mainlytriacylglycerols(TAGs)with minoramountsofothercompoundssuchasfreefattyacids, mono-anddiacylglycerols[26,27].Withthepurposeofreducingthesize of thekineticsmodellingtask,agenerallyacceptedapproach is toexemplifythevegetableoilexaminedeitherbyoneTAGonly [28–30], or as a binary mixtureof triolein and oleic acid [31]. Recently,therewereattemptstorepresentsomevegetableoilsasa mixtureofseveralTAGs,withavariedsuccess—fromafailurein thepredictionof thephase equilibriumof themulticomponent mixtureexamined[32]toanacceptablequantitativeand qualita-tiverepresentationofthekineticcurvesmeasured[33].

Inthecaseofchiaseedoil,ouranalysesshowthatlinolenicacid isthefattyacidwiththehighestcontentfollowedbylinoleicacid. Hence, the TAGs trilinolenin and trilinolein are themain lipid representativesintheoil.

Tosimulatetheextractionkineticsofoilfromthechiaseeds,the phasebehaviorofthesystem(oilandscCO2)shouldbemodeled,

which requiresanappropriatethermodynamicmodelbywhich

the solubility of theoil in the scCO2 will be calculated. Then,

followingthealgorithmadvocatedinSovovaand Stateva[21],a secondorderpolynomialfunctionwillbefittedtothesolubility

dataandimplementedinthedynamicmodel.

Generally, equations of state(EoSs) are theusualchoice for calculationofsolubilityofacompound(mixtureofcompounds)in scCO2. Their application requires knowledge of the critical

temperature and pressure of the pure compounds comprising

themixture.It shouldbenoted,however,thatinmanycasesof

complex systems, which, for different reasons, have to be

represented by a model compound(s), not always the most

appropriateone(s)ischosen,becauseofthelackofinformation onits(their)criticalproperties.

Inourcase, themostsuitablerepresentativeofchiaseedoil, followingtheresultsoftheanalyses,istheTAGtrilinolenin.Having saidthat,however,twoveryimportantissuesshouldbetakeninto consideration:i)lackofanyexperimentalinformationontheVLE oftrilinolenin+scCO2;ii)totallackofdata(bothexperimentaland

estimated)onthethermophysicalpropertiesoftrilinolenin. Hence,thoughtrilinoleninisthemostadequaterepresentative ofthechiaseedoil,theuncertaintiesthatwillbeintertwinedinto thesolubilitypredictionsviaitsestimatedpropertiescouldbeso substantialthattheymightleadtoamisrepresentationoftheVLE ofthebinary(trilinolenin+scCO2),e.g.toanerroneousprediction

oftheextentofthevapour-liquidregion,which,however,could notbeidentifiedandverified(seeissuei).

Inviewoftheabove,wechosetrioleinastheTAGtorepresent thechiaseedoil,whichwasmotivatedbytworeasons:1)there

are experimental data available on the VLE of the binary

of triolein used by us have been proved to represent in an acceptablewaytheVLEofthesystem[33].

Thesolubility(molefraction)ofacompoundintheSCfluidmay beexpressedas: yi¼ xi

’

Li

’

V i ð10Þ where

’

L

i and

’

Vi are the fugacity coefficients of triolein, representingthechiaoil inourcase, intheliquid andSC fluid phase,respectively.

WeemploythepredictiveSoave-Redlich-Kwong(PSRK)cubic

EoS[34] to calculatethefugacity coefficients of triolein inthe liquidandvaporphases,respectively,andusethevaluesforits criticaltemperatureandpressurereportedbyCoelhoetal.[33]. Resultsanddiscussion

SupercriticalCO2extraction

Theexperimentalkineticcurves,obtainedfortheSCEofchiaoil fromDCSandECSaredisplayedonFig.1aandb,respectively.For DCS,the extractionyield (massof oil/mass of seeds)increased withpressureandtemperature,andwasintherangefrom13.3% (at25MPaand40
C)to18.6%(at45MPaand60
C)after240min extractiontime(Fig.1a).

Attheexperimentalconditionsstudiedinthiswork,acrossover effectontheoverallextractionyieldwasnotobserved.RochaUribe

et al. [19] at operational conditions (27.2–40.8)MPa and

(40–60)o_{C, which are verysimilar to ours, reportedanalogous} behaviorpattern, while Ixtainaetal. [18] observed a crossover

point within the same range of pressures (25–45)MPa and

temperatures(40–60)o_{Casthosestudiedbyus.}

Considering thetotal oil (19.9%mass)contained inDCS,the recoveryvalues (massofoil extracted/massof oilin theseeds)

achieved in this work ranged from (66.7–93.5) %, and are

consonantwiththeresultsreportedbyIxtainaetal.[17,18]and

Scapin et al. [20], who employed high-quality chia seeds

containing (32–34) % mass of oil, from different geographical origins,asarawmaterial.

Fig.1bshowstheCO2flowrateeffectontheoilyieldwhenECS

areused.Pressureandtemperatureweresetat45MPaand40
C,

respectively. The reasons behind choosing these particular

experimentalconditionswerethattheyprovedtobetheoptimal onesforobtaininghigheroil recoveriesandALAconcentrations fromDCS(seeSection“Analysisoffattyacidcomposition”).

Asshown(Fig.1b),extractioncurvesoverlapattheendofthe extractionprocess, and hence similar oil extractionyields (24.6–25.2)% wereachievedindependentlyoftheCO2flowrateapplied.Takinginto

considerationthattheinitialoilcontentforECSwas28.3%mass,and thattherecoveriesobtainedwerebetween(86.9–89.9)%,itcanbe concludedthat practicallyall availableoilwas extractedafterthe extractiontime(240min).Nevertheless,duringtheearlystagesof theextraction,whenthefreeoillocatedonthesurfaceoftheseedsis extractedandthemasstransferresistanceisnegligible,theextraction rateincreasedwiththe CO2flow.Atthatpoint,themassofoilextracted

atthelowestCO2flowrate(0.44gmin1)was1.7-foldhigherthanthe

obtainedatthehighestflowrate(0.76gmin1). Analysisoffattyacidcomposition

Fig.2showsthefattyacidcompositionofoilsfrombothsetsof chiaseeds.ALAwasthemainfattyacid(55.58–67.45%)intheoils, followedbylinoleicacid(17.2–23.5%).InrespectofDCS(Fig.2a),

Fig.1.Experimental(symbols)andsimulated(lines)cumulativeextractioncurvesobtainedfor(a)DCS(40gmin1CO2flowrate),and(b)ECS(45MPaand40
C).

Fig.2. Fattyacidcomposition(%oftotalfattyacids)ofoilsextractedfrom(a)DCS (40gmin1CO2flowrate)and(b)ECS(45MPaand40
C).Extractiontime:240min.

thefattyacidprofilewasverysimilarforbothpressuresexamined (25and45MPa).Withregardtotheinfluenceoftemperature,the concentrationofALAintheextractwas lowerat60
C (n-6/n-3 ratiosaround0.26and0.42at40and60
C,respectively,datanot shown).Likewise,thefattyacidprofilesobtainedfromECSatthe differentCO2flowratios(Fig.2b)wereverysimilar(PUFAandALA

concentrations around (81%and 59%, respectively). Our results agreewellwiththoseofotherauthors,whousedchiaseedsfrom several geographical origins, and applied different extraction methodsandoperationalconditions[7–11,17–20].

Kineticsmodellingresultsanddiscussion

ThemodeldescribedinSection“Supercriticalfluidextractionof

chia oil — modelling framework” was solved in gPROMS to

simulatetheevolutionofyieldovertimefortheoilextractedfrom thetwodifferentsetsofchiaseeds,namelyDCSandECS,atthe operationalconditionsofinteresttotheexperiment.

For the ECS, the initial oil content of the matrix, wsum, is 0.283kgkg1 solid, while for DCS wsum;0 is 0.199kgkg1 as discussedin Section“Supercriticalfluid extractionof chiaoil — modellingframework”.

Therearefourunknownparametersinthemodel:b,kf,wtandK.

Asthenumberofexperimentaldatapointslimitsthenumberof

model parameters that can be estimated within a reasonable

confidenceinterval,predefinedfixedvaluesshouldbeassignedto someoftheaboveparameters.

Thevalueofparameterb,whichshouldbe»1,wasdetermined aftersomepreliminarycalculationsandsensitivityanalysisofits influenceontheextractionkineticsmodelling.Itwasverifiedthat thebestvaluewasb=40, andhenceitwasusedinthekinetics modelforallcasesofDCSandECSexamined.

Thevalueofkfforeachcasestudiedwasestimatedfollowing

Coelhoetal.[33],whousedtherelationofWilkeandChang[35]. Thekfvaluesobtainedwerethensetasconstantsinthekinetics

model,thusreducingthedegreesoffreedom.FortheDCScase,the bestvaluesofkfobtainedaredisplayedinTable1.FortheECScase,

wherepressureandtemperatureremainconstant,asinglekfvalue

wasusedforthethreescCO2flowratesapplied(Table2).

Hence, the maximum content corresponding to monolayer

adsorption,wt,andthepartitioncoefficientKarethetwomodel

parameterslefttobeestimatedbyfittingthedynamicmodeltothe experimentaldata.

ThebestestimatedvaluesofKandwtareshowninTables1and2,

fortheDCSandECScases,respectively.

For DCS, the influence of temperature on the partition

coefficientispronounced—Kincreasesbyanorderofmagnitude withtheincreaseoftemperature,whiletheincreaseofKvalues withpressureisnotsonoticeable(Table1).Thisbehaviorshows

thatthebondbetweenthesoluteandthematrixweakenswith

increasingtemperatureofextraction,which,inturn,leadstoan increase in partition coefficient values, favoring thus the CO2

phase,whiletheadsorbentcapacitygenerallydecreases. TheinfluenceofscCO2flowrateonKisdemonstratedforthe

caseofSCEofECS(Table2).Asshown,thevaluesofKdecreasewith theincreaseofscCO2flowrate.

Thedeviationsbetweentheexperimentallymeasuredandthe calculated yields, expressed by the AARDs (%), are also shown (Tables1and2,respectively).For theDCScase, thereisa good qualitativeandquantitativeagreementbetweenthesimulatedand experimentalextractionyieldcurves,asdemonstratedonFig.1a

Table1

EstimatedvaluesofKandwt,andmasstransfercoefficients(kf)fortriolein,atdifferentexperimentalconditionsand constantCO2 flowrate(40gmin1)fromDCS.

The AARDsrepresent the deviations betweenthe experimental and calculated yieldvalues. P(MPa) T(
_{C) K(kgplantkg}1_CO

2) kf(min1) wt(kgkg1CO2) AARD(%)

25 40 3.22E-3 1.51E-5 8.04E-2 4.82

25 60 1.36E-2 3.08E-5 8.04E-2 9.49

45 40 4.27E-3 7.03E-4 5.76E-2 5.64

45 60 2.47E-2 4.58E-4 5.20E-2 2.13

Table2

EstimatedvaluesofKandwtfortrioleinat45MPaand40
C,andvaryingscCO2flow

rateforECS.Trioleinmasstransfercoefficientskf=7.03E-4.TheAARDs represent

deviations between the experimental and calculatedyield values. F(kgmin1_{) K(kgplantkg}1_CO

2) wt(kgkg1CO2) AARD(%)

27 2.00E-2 6.01E-2 1.22 40 1.69E-2 8.25E-2 2.47 54 9.28E-3 8.07E-2 6.65

Fig.3.Simulatedprofileofoilconcentrationinthesolid(chiaseeds,largepicture)andfluid(scCO2,smallpicture)phasesduringextractionfor(a)DCS(40gmin1CO2flow

andTable1.Theonlyexceptionbeingthecaseat25MPaand60
C wheretheAARDishigher.Thelatterwasnotunexpectedtaking intoconsiderationthesomewhatdifferenttrendoftheparticular experimentalextractioncurveincomparisontotheotherthree.

FortheECScase,thesimulatedextractioncurvesfollowvery wellthepatternoftheexperimentalonesandoverlapattheendof the extraction. Thus, the kinetics modelling results verify the experimentallyobservedfactthattheoilextractionyieldsarenot influencedbytheCO2flowrateapplied.TheAARDvaluesobtained

confirmtheverygoodagreementbetweentheexperimentaland

simulatedresults(Table2).

Theevolutionoftheoilconcentrationduringextractioninthe solidmatrixinsidetheextractorandintheexitingfluidstreamwas simulated.The resultsforDCSand ECSarepresentedinFig.3a andb,respectively.

AsdepictedinFig.3a,increasingthepressureleadstoafaster andmoreefficientextraction.However,thateffectismoderateat 40
Candatthefirststageoftheextraction,whileitbecomesmore significantwhentheoperatingtemperatureissetto60
C.

Fig.3bshowsapositiveeffectoftheincreasedscCO2flowrate

onthespeed ofextraction.Thus, forECS, ontheonehand,the changeofoilconcentrationinthesolidphasefollowsthepattern observedfortheDCS(Fig.1a),andreachesthesamefinalvalue regardlessoftheCO2flowrate.Ontheotherhand,ahigheryield

and lower solid phase concentration are being achieved with

smallerflowrates.

Asfarasweareaware,thesearethefirstdatademonstrating theevolvementoftheoilconcentrationinboththefluidandsolid

phases during scCO2 extraction of chia seeds. Taking into

considerationthatasinglemodelcompoundisusedtorepresent theoil,theresultsobtainedarequiteadequate.Furthermore,they

provide valuable information to be used with confidence in a

subsequentprocessdesignstage. Conclusions

Thisworkpresentsforthefirsttimetheresultsofmodellingthe experimentalkineticsdataofSCEofoilfromtwosetsofchiaseeds, ECSandDCS.TheSCEexperimentsdemonstratedthatthehighestoil yield(18.6%)obtainedfromDCS was achieved atthehighest pressure andtemperatureapplied(45MPaand60
C).Furthermore,atthese operationalconditionspracticallyalltheoilwasexhausted(93.5%oil recovery).Ascanbeexpected,theextractionyieldsachievedfrom ECS,ascomparedtothosefromDCS,werehigher(24.6–25.2)%,but their values were not influenced by theCO2 flow rate applied.

Nevertheless,theincreaseintheCO2-to-chiamass ratioenhanced up

to1.7timestheoilextractionrateattheearlystagesofextraction. ConcentrationsofALAintherange(55–67)%inoilswereattained.

Furthermore,the oils obtained from both seeds (DCS and ECS)

presentedasimilarfattyacidprofile.

The kinetics modelling was performed applying a new

approach, which intertwines the complex interaction between

kineticsandsolubility.Forthepurposeofmodelling,trioleinwas

chosen as chia seeds oil representative compound. The model

equationsweresolvedingPROMSModelBuilderenvironment,and

parameter estimation was performed to obtain some model

parameters.

Theresultsobtaineddemonstratethatalbeitthesimplifications introducedinthemodel,thereisagoodagreementbetweenthe calculatedandexperimentalextractionyieldsattheSCEoperating conditionsexamined.

Finally,thevaluableinformation onthemasstransferofthe extractionprocessofECSandDCSobtainedcanserveasa solid basisforthedevelopmentofindustrialapplicationstargetingthe valorizationandmonetarizationofchiaseedsandinparticularof thehighlyunderusedDCS.

Declarationsofinterest None.

Acknowledgements

J.A.P.Coelho,R.M.FilipeandR.P.Statevaacknowledgethe

funding received from the European Union’s Horizon 2020

researchand innovationprogramme underthe Marie

Sklodow-ska-CuriegrantagreementNo.778168.J.A.P.Coelho,andR.M.

Filipe acknowledge thefundingreceived fromFundaçãopara a

Ciência ea Tecnologia,Portugal, underprojectsUID/ECI/04028/

2013 and UID/QUI/00100/2013. Chia seeds were generously

suppliedbyPRIMARIA(www.primaria.biz). References

[1]M.I.Capitani,S.M.Nolasco,M.C.Tomas,FoodHydrocoll.61(2016)537,doi: http://dx.doi.org/10.1016/j.foodhyd.2016.06.008.

[2]B.L.Olivos-Lugo,M.A.Valdivia-Lopez,A.Tecante,FoodSci.Technol.Int.16(1) (2010)89,doi:http://dx.doi.org/10.1177/1082013209353087.

[3]E.Reyes-Caudillo,A.Tecante,M.A.Valdivia-López,FoodChem.107(2)(2008) 656,doi:http://dx.doi.org/10.1016/j.foodchem.2007.08.062.

[4]P.Porras-Loaiza,M.T.Jiménez-Munguía,M.E.Sosa-Morales,E.Palou,A. Lopez-Malo,Int.J.FoodSci.Technol.49(2014)571,doi:http://dx.doi.org/10.1111/ ijfs.12339.

[5]G.Barceló-Coblijn,E.J.Murphy,Progr.LipidRes.48(6)(2009)355,doi:http:// dx.doi.org/10.1016/j.plipres.2009.07.002.

[6]C.Galli,C.F.Marangoni,Prostag.Leukotr.Ess.FattyAcids75(2006)129,doi: http://dx.doi.org/10.1016/j.plefa.2006.05.007.

[7]G.Dąbrowski,I.Konopka,S.Czaplicki,M.Tanska,Eur.J.LipidSci.Technol.119 (2017)1,doi:http://dx.doi.org/10.1002/ejlt.201600209.

[8]B.T.F.deMello,V.A.dosSantosGarcía,C.daSilva,J.FoodProcessEng.40(2015) 1,doi:http://dx.doi.org/10.1111/jfpe.12298.

[9]M.R.Segura-Campos,N.Ciau-Solís,G. Rosado-Rubio,L.Chel-Guerrero,D. Betancur-Ancona, Agric. Sci. 5 (2014) 220, doi:http://dx.doi.org/10.4236/ as.2014.53025.

[10]R.G.Tolentino,M.L.R.Vega,S.VegayLeón,J.Fontecha,L.M.Rodríguez,A.E. Medina,Rev.Cub.Plant.Med.19(2014)199.

[11]A.B.Zanqui,D.R.deMorais,C.M.daSilva,J.M.Santos,L.U.R.Chiavelli,P.R.S. Bittencourt,M.N.Eberlin,J.V.Visentainer,L.Cardozo-Filho,M.Matsushita,J. Brazil Chem. Soc. 26 (2) (2015) 282, doi:http://dx.doi.org/10.5935/0103-5053.20140278.

[12]G. Brunner, J. Food Eng. 67 (2005) 21, doi:http://dx.doi.org/10.1016/j. jfoodeng.2004.05.060.

[13]M.Wei,S.Hong,Y.Yang,J.Ind.Eng.Chem.48(2017)202,doi:http://dx.doi.org/ 10.1016/j.jiec.2017.01.003.

[14]J.N.Moon,A.T.Getachew,A.S.M.T.Haque,P.S.Saravana,Y.J.Cho,D.Nkurunziza, B.S.Chun,J.Ind.Eng.Chem.69(2019)217,doi:http://dx.doi.org/10.1016/j. jiec.2018.09.019.

[15]M.Topiar,M.Sajfrtova,J.Karban,H.Sovova,J.Ind.Eng.Chem.77(2019)223, doi:http://dx.doi.org/10.1016/j.jiec.2019.04.041.

[16]O.Benito-Román,S.Varona,M.T.Sanz,S.Beltrán,J.Ind.Eng.Chem.80(2019) 273,doi:http://dx.doi.org/10.1016/j.jiec.2019.08.005.

[17]V.Y.Ixtaina,A.Vega,S.M.Nolasco,M.C.Tomás,M.Gimeno,E.Bárzana,A. Tecante, J. Supercrit.Fluids 55(2010) 192, doi:http://dx.doi.org/10.1016/j. supflu.2010.06.003.

[18]V.Y.Ixtaina,F.Mattea,D.A.Cardarelli,M.A.Mattea,S.M.Nolasco,M.C.Tomás,J. Am.OilChem.Soc.88(2011)289, doi:http://dx.doi.org/10.1007/s11746-010-1670-2.

[19]J.A.RochaUribe,J.I.NoveloPérez,H.CastilloKauil,G.RosadoRubio,C.G. Alcocer, J. Supercrit. Fluids 56 (2011) 174, doi:http://dx.doi.org/10.1016/j. supflu.2010.12.007.

[20]G.Scapin,E.R.Abaide,L.F.Nunes,M.A.Mazutti,R.G.Vendruscolo,R.Wagner,C. S.daRosa,J.Supercrit.Fluids127(2017)90,doi:http://dx.doi.org/10.1016/j. supflu.2017.03.030.

[21]H.Sovova,R.P.Stateva,Ind.Eng.Chem.Res.54(2015)4861,doi:http://dx.doi. org/10.1021/acs.iecr.5b00741.

[22]M.P. Castro-Gomez, L.M. Rodriguez-Alcalá, M.V. Calvo, J. Romero, J.A. Mendiola,E.Ibanez,J.Fontecha,J.DairySci.97(2014)6719,doi:http://dx. doi.org/10.3168/jds.2014-8128.

[23]D.Villanueva-Bermejo,F.Zahran,M.R.García-Risco,G.Reglero,T.Fornari,J. Supercrit. Fluids 119 (2017) 283, doi:http://dx.doi.org/10.1016/j.sup-flu.2016.10.005.

[24]P.Castro-Gómez,J.Fontecha,L.M.Rodríguez-Alcalá,Talanta128(2014)518, doi:http://dx.doi.org/10.1016/j.talanta.2014.05.051.

[25]ProcessSystemsEnterprise,gPROMS.https://www.psenterprise.com/gproms/

.(Accessed10November2018).

[26]V.Y.Ixtaina,M.L.Martínez,V.Spotorno,C.M.Mateo,D.M.Maestri,B.W.K.Diehl, S.M.Nolasco,M.C.Tomás,J.FoodCompos.Anal.24(2011)166,doi:http://dx. doi.org/10.1016/j.jfca.2010.08.006.

[27] Y.P.Timilsena,J.Vongsvivut,R.Adhikari,B.Adhikari,FoodChem.228(2017) 394,doi:http://dx.doi.org/10.1016/j.foodchem.2017.02.021.

[28]K.Araus,E.Uquiche,J.M.delValle,J.FoodEng.92(2009)438,doi:http://dx.doi. org/10.1016/j.jfoodeng.2008.12.016.

[29]R.M.Couto,J.Fernandes,M.D.R.G.daSilva,P.C.Simões,J.Supercrit.Fluids51 (2009)159,doi:http://dx.doi.org/10.1016/j.supflu.2009.09.009.

[30]K.S.Duba,L.Fiori,J.Chem.Thermodyn.100(2016)44,doi:http://dx.doi.org/ 10.1016/j.jct.2016.04.010.

[31] I.Gracia,M.T.García,J.F.Rodríguez,M.P.Fernández,A.deLucas,J.Supercrit. Fluids48(2009)189,doi:http://dx.doi.org/10.1016/j.supflu.2008.11.006.

[32]H.P.Cornelio-Santiago,C.B.Gonçalves,N.A.deOliveira,A.L.deOliveira,J. Supercrit. Fluids 128 (2017) 386, doi:http://dx.doi.org/10.1016/j.sup-flu.2017.05.030.

[33]J.P.Coelho,R.M.Filipe,M.P.Robalo,R.P.Stateva,J.Supercrit.Fluids14(2018)68, doi:http://dx.doi.org/10.1016/j.supflu.2017.12.008.

[34]T.Holderbaum,J.Gmehling,FluidPhaseEquilibr.70(1991)251,doi:http://dx. doi.org/10.1016/0378-3812(91)85038-V.

[35]C.R.Wilke,P. Chang,AIChEJ.1(1955)264,doi:http://dx.doi.org/10.1002/ aic.690010222.