Mineral oil hydrocarbons are used in the consumer goods sector for the elaboration of a wide range of foods and cosmetics. Traditional methods for determining their levels and composition are time consuming and laborious, besides requiring complex instrumentation.
Trang 1journalhomepage:www.elsevier.com/locate/chroma
Alan Rodrigo García-Cicourela,∗, Bas van de Veldea, Gerry Roskama, Hans-Gerd Janssena,b,c
a University of Amsterdam, Van‘t Hoff Institute for Molecular Sciences, Analytical Chemistry Group, P.O Box 94157, 1090 GD Amsterdam, the Netherlands
b Wageningen University, Agrotechnology and Food Sciences Group, Laboratory of Organic Chemistry, Stippeneng 4, 6708 WE Wageningen, the Netherlands
c Unilever Research and Development, P.O Box 114, 3130 AC Vlaardingen, the Netherlands
a r t i c l e i n f o
Article history:
Received 6 September 2019
Revised 11 November 2019
Accepted 14 November 2019
Available online 15 November 2019
Keywords:
Mineral oil analysis
Supercritical fluid chromatography
Argentation chromatography
Parallel detection
MOSH/MOAH deconvolution
a b s t r a c t
Mineraloilhydrocarbonsareusedintheconsumergoodssectorfortheelaborationofawiderangeof foodsandcosmetics.Traditionalmethodsfordeterminingtheirlevelsandcompositionaretime consum-ingand laborious,besidesrequiringcomplexinstrumentation Hereasimpleand fastmethodwas de-velopedthatusescolumnspackedwithsilver-modifiedsilicainsupercriticalfluidchromatographywith flameionizationandUVdetection(SFC-FID/UV)forthedeterminationofmineraloilsaturated hydrocar-bons(MOSH)andmineraloilaromatichydrocarbons(MOAH)inpurifiedmineraloilsamples.Themethod requiresnosamplepreparationapartfromdilution.DirectquantificationofMOSHandMOAHwas pos-sibleforsamples withMOSH/MOAHratiosaround one.Forothersamples deconvolutionofthe MOSH andMOAHhumpsintheFIDchromatogramusingtheUVsignalwasneededsincebaselineseparation
ofthetwofractionscouldnotbeobtained.Validationofthemethodperformanceshowedanexcellent linearity(R2>0.9995)intherangeofconcentrationstested(2.5–100mgmL−1)andabetterrepeatability thanthestandardmethods (<5%).MOAHdetectionlimitswerebetterthan0.36%MOAH,whichmakes the methodsufficientlysensitive for analysis ofallbut the highest purity mineral oils.The proposed SFC-FID/UVmethodwassuitablefortheanalysisofmineraloilswithviscositiesandmolecularweights belowapproximately56mm2s−1 and450gmol−1.Thequantitativeresultsofthenewmethodwerenot statisticallysignificantlydifferentfromthoseobtainedwiththestandardSPE-GC-FIDmethodwherethe newmethodhastheadvantagesofabetterrepeatability,simpleroperationandasignificantlyshortened analysistime.Thisnewmethodcouldpotentiallyalsobeusedfortheanalysisofmineraloil contamina-tionsinconsumerproductssuchasfoods.However,inthiscaseadditionalsampleclean-upand precon-centrationstepsareneededforreducingmatrixinterferencesfrome.g.triglyceridesandolefinsandfor improvingthedetectionlimits
© 2019TheAuthors.PublishedbyElsevierB.V ThisisanopenaccessarticleundertheCCBY-NC-NDlicense
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
1 Introduction
Thepresenceofmineraloilhydrocarbonsinfoodsand
cosmet-icsiscurrentlyreceivingagreatdealofattentionintheconsumer
goods sector Mineral oil hydrocarbons can enter these products
non-intentionally asa contaminant from e.g packaging or
trans-port,butpurifiedmineraloilfractionsarealsowidelyusedas
in-gredients for the elaboration of a range of foods and cosmetics
Detailedanalysesareneededtoestablishthelevelofresidual
aro-∗ Corresponding author
E-mail address: a.r.garciacicourel@uva.nl (A.R García-Cicourel)
maticspeciesinthepurifiedoilsandtodeterminetheextentand typeofamineraloilcontamination.Inbothapplicationareas sep-arationofthemineraloilsaturatedhydrocarbons(MOSH)fromthe mineraloilaromatichydrocarbons(MOAH)isessential.Several liq-uid chromatographic methods to do so have been described [1– 5].Combined with auniversal detector, generallya flame ioniza-tiondetector(FID), theseparatedMOSH andMOAH fractionscan
be quantified In general, however, these methods are time con-suming,laboriousandrequiredifficulttooperatetwo-dimensional heart-cutLC-GCinstruments[6]
Separation of complex hydrocarbon mixtures based on aro-maticity(polarity) is difficult because it requiresa very high se-https://doi.org/10.1016/j.chroma.2019.460713
0021-9673/© 2019 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )
Trang 2lectivity towards the aromatic ring system of the molecules [7].
Whereasgaschromatography(GC) canprovide thisselectivity for
very small molecules,GC selectivity for the very high molecular
weight compounds encountered in typical MOSH/MOAH analyses
isinsufficient Still, GC is widely used as a second dimension in
MOSH/MOAH analyses because of its easy coupling to the FID,
whichis often preferreddueto its equal response forall
hydro-carbons irrespective of their structure [8,9] LC stationary phases
ontheotherhandcanprovideagoodselectivityforMOSH/MOAH
fractionation[1,7,8],butunfortunatelyLC systemscannot be
cou-pledto the FID MOSH/MOAH analyses hencerequire coupledLC
andGCmethods:LCusingnormalphasecolumns(NPLC)for
frac-tionationofthemineraloil basedonaromaticity,followedby GC
analysisusingeitherFID forquantification,ormass spectrometry
(MS)orvacuumultraviolet detection(VUV) forstructural
charac-terization [6,8,10] This combined LC-GC approach was first
de-scribedby Biedermann etal in2009 andhas beenfurther
opti-mized and improved in a series of articles by the same authors
andotherusersofthemethod[11–14].Asignificantimprovement
wasthe use of silver-loaded silica columns which made the
de-terminationofthecut-pointforthe MOSH/MOAHseparation less
critical[10,13].Barp etal.optimizedtheoriginal Biedermannand
GrobmethodbyspeedinguptheLCandGCanalyses.Ananalysis
timeof23minwasachievedfortheanalysisofeachfraction,
re-sultingin a total analysistime of 46 minwhen both MOSH and
MOAHlevels havetobe determined [12].Despitethese
improve-ments,thelonganalysistime andthecomplicatedcoupledLC-GC
set-uprequiredstill maketheapplicationofthetwo-step method
inroutineanalysisexpensiveandcomplex
In many applications supercritical fluid chromatography (SFC)
has been shown to combine the advantages of LC and GC [15]
The technique, for example, is used in the mineral oil industry
ina numberofmethods fortheanalysis oflighter fuels Method
D5186–19 from the American Society for Testing and Materials
(ASTM),asanexample,describesSFCwithpackedcolumnsasan
officialmethodforthedeterminationofthearomaticand
polynu-cleararomatic content of light diesel- and aviation fuels [16] In
SFC,allLCstationaryphasescanbeused.Silicacolumnshavebeen
extensivelyapplied[17],andalsosilver-loaded silicacolumnscan
beused[18].Ifthesecolumnswouldbecapableofseparatingthe
heaviermineraloilsintoMOSHandMOAHusingpurecarbon
diox-ide as the mobile phase, MOSH and MOAH levels could be
de-termineddirectly using justone singlechromatographic method,
SFC-FID.Additionally,SFCcould alsosolve other issuesrelatedto
theonlineLC-GCmethod,such asincompleteelutionof thehigh
molecularweightcompoundsintheGC-FIDquantificationstep
In thiscontribution,theapplicabilityofpacked-column SFCas
afasterandsimple methodforthe analysisofMOSHandMOAH
incrudeandpurifiedmineraloilsamplesisassessed.Different
sta-tionaryphasesareevaluatedfortheirabilitytoseparateMOSHand
MOAHand provide informationon the aromaticity of theMOAH
species.Parallel detection using UV and FID is studied to
maxi-mize sensitivityand selectivity of the method Pure CO2 is used
asthe mobile phase in all experiments to allow couplingto the
FID.The applicability andmeritsof the newSFC-FID/UV method
are discussed,and quantitative dataobtained are compared with
thosefromthestandardoff-line solid phaseextraction–
SPE-GC-FIDmethodforMOSH/MOAHanalysis
2 Materials and methods
Standards to study the selectivity in SFC were purchased
from Sigma-Aldrich (Zwijndrecht, the Netherlands) These
comprise the aliphatic hydrocarbons n-undecane (C11),
bicy-clohexyl (Cycy), n-tridecane (C13) and 5-α-cholestane (Cho),
the mono-aromatics 1,3,5-tri-tert-butylbenzene (TBB) and
1,4-di-dodecylbenzene (D12B), and finally the poly-aromatics 1-methylnaphthalene (1MN), biphenyl (BP) and anthracene (Ant) Hexane and dichloromethane (DCM), both HPLC grade, were purchased from Biosolve BV (Valkenswaard, the Netherlands) SFC grade carbon dioxide (grade 4.8) wasobtained from Praxair (Vlaardingen, the Netherlands) Silver-nitrate impregnated silica gel(230 mesh– 65μm) loaded atapproximately10wt.%forthe SPEexperimentswaspurchasedfromSigma-Aldrich.Eightmineral oilsamples fromdifferent stagesinthepurification process were obtainedfroma localsupplierof whiteoils.The physicochemical properties of the mineral oils are summarized in Table 1 Stock solutions ofthe standard compounds (at10 mgmL−1) andofthe mineraloils(at500mgmL−1)werepreparedinhexaneandstored
intherefrigeratorwhennotinuse
2.1 SFC-FID/UV selectivity
For the experiments to map the SFC selectivity, mixtures of aliphaticmarkercompounds(C11,C13,CycyandCho)andselected aromaticstandards(1MN,BP,TBB,D12BandAnt)werepreparedin hexaneatlevelsaround300μgmL−1percompound.Separationof thecompoundswasperformedonanAcquityUPC2SFCinstrument withPDAdetection(Waters,Etten-Leur, theNetherlands) coupled
toanexternalFID.ForFIDdetection,partoftheCO2mobilephase flow was diverted to the FID To do so, a low dead-volume T-piece(ValcoInstrumentCompanyInc.,Schenkon,Switzerland)was connected to the exit ofthe column just prior to the UV detec-tor A 90cm long, 50μm internal diameter fused-silicacapillary wasusedtotransferthemobilephaseflowtotheFIDofan adja-cent Agilent G1530 GC system (Agilent, Amstelveen, the Nether-lands) At the end of the capillary an integral tapered restrictor waspreparedusingthemethoddescribedbyGuthrieandSchwartz [19].Inshort,thecapillarywasfirstsealedby heatingitina bu-tane/nitrousoxide flameandthen re-opening itby gently abrad-ingitbyhandusingawetabrasivesheet.There-openingisdone withthecapillaryconnectedtotheSFCpump.Gasbubbles escap-ingfromthecapillaryindicatetherestrictorisopen.Abradingwas continueduntila gaseousCO2 flowofapproximately20mLmin−1 wasobtained The outlet ofthe transfer capillarywaspositioned justbelow the flame jet of the FID The temperatures of theGC andFIDweresetat60and300°C,respectively.Hydrogenandair flowsfortheFIDwere36and400mLmin−1,respectively
TheinjectionvolumeintheSFCexperimentswas1μL.Forthe separation ofthe compounds a columnset consistingofeither a baresilicacolumn,150× 4.6mmPolaris5Si-A5μm,ortwo se-rially connected 100 × 4.6 mm silver-loaded silica Chromsep SS
5μmcolumns(Agilent)wasused.PureCO2 wasusedasthe mo-bilephaseataflowrateof1mLmin−1.Temperatureandpressure were set at60 °Cand110 bar, respectively.Data acquisition was performed at 20Hz forthe FID and 25 Hz for the UV detector
UV chromatograms were acquiredand processedusing Empower
3(Waters) eitherinthe MaxPlot mode (190–400 nm)for decon-volution,orat254nmformaximumselectivitytowardsaromatic species
Thesamemixtures ofaliphaticandaromaticcompounds were analyzedbyLConaShimadzu10AvpHPLCsystemequippedwith
UV and refractive index (RID) detection (Shimadzu Benelux, ‘s-Hertogenbosch, the Netherlands) To do so, 20 μL of each mix-turewere injectedontothetwoseriallyconnected100× 4.6mm silver-loadedsilicacolumnspreviouslydescribed.Compoundswere elutedusinghexaneasthemobilephaseat0.5mLmin−1
2.2 Optimization of the SFC-FID/UV method for mineral oil analysis
Optimization ofthe SFCparameters, i.e.temperature, pressure andflow rate, wasdone using a two-level factorial experimental
Trang 3Table 1
Physicochemical properties of seven mineral oil samples
Density (gcm −3 ) Viscosity at 40 °C (mm 2 s −1 ) Molecular weight (gmol −1 ) Number of carbon atoms Mineral oil 1 0.8570 16.53 352.0 C 15 –C 40 , centered around C 24
Mineral oil 2 0.8631 10.32 284.0 C 14 –C 31 , centered around C 23
Mineral oil 3 0.8694 30.29 379.0 C 14 –C 40 , centered around C 25
Mineral oil 4 0.8984 22.56 312.0 C 13 –C 40 , centered around C 23
Mineral oil 5 0.8742 56.41 452.0 C 16 –C 44 , centered around C 28
Mineral oil 6 0.8859 91.21 519.0 C 17 –C 52 , centered around C 33
Mineral oil 7 0.9158 153.40 423.0 C 18 –C 48 , centered around C 30
Mineral oil 8 0.8284 18.62 388.0 C 14 –C 31 , centered around C 23
Table 2
High and low levels of the selected parameters
for the optimization of the SFC-FID/UV method
by a two level factorial experimental design
Parameter Low level High level
Temperature ( °C) 50 60
Pressure (bar) 110 138
Flow (mLmin −1 ) 1.0 2.0
design.Thehighandlowlevelsforthethreeparametersareshown
inTable2.TheresponsetooptimizewastheMOSH/MOAH
resolu-tion.Intheseexperiments1μLofmineraloil1wasinjectedata
concentrationof10mgmL−1.Forthecalculationoftheresolution,
retention timesand peak widthsof the MOSHand MOAHbands
were calculatedboth fromthe FIDandUV chromatograms.These
datawerealsousedtoassessextra-columnbandbroadeninginthe
capillarytotheFID
2.3 Validation of the SFC-FID/UV method
To evaluate the performance of the proposed SFC-FID/UV
method,linearity,repeatability, limitofdetection (LOD)andlimit
of quantification (LOQ) were assessed following the Eurachem
guideline formethodvalidation[20].The experimentswere done
using solutions of mineral oil 8 in hexane This oil was
aromat-icsfree.Linearitywasevaluatedintheconcentrationrangeof2.5–
100 mgmL−1 Repeatability wasevaluated at three concentration
levels, 7.5, 15 and30 mgmL−1 LOD andLOQ were calculated by
injectingthreedifferentsolutionsofmineraloil8at0.25mgmL−1
(five times each).For all validationexperiments, 1 μLof the
dif-ferent sampleswere injected.Theseexperiments were performed
at60°C and138barat asupercritical CO2 mobilephase flow of
1 mLmin−1 The other instrumental parameters were maintained
asmentionedintheSection2.1
2.4 Analysis of real samples
Solutions ofmineral oils 1–7atconcentrations of10 mgmL−1
in hexane were analyzed by SFC-FID/UV The injection volume
was1 μL.CO2 temperature, pressureandflow were setat60 °C,
138 bar and1 mLmin−1, respectively All other settings were as
indicated in Section 2.1 To allow comparisonof the quantitative
dataobtainedusingSFC-FID/UVwiththoseofthestandard
meth-ods, also off-line SPE-GC-FID analyses were performedaccording
to theprocedure describedpreviously [6].The GC-FIDanalysisof
the isolatedMOSH andMOAHfractions wascarried out usingan
Agilent 6890 Ninstrument equipped witha Focus-PAL
autosam-pler(GLSciences,Eindhoven,theNetherlands) Thesample(1μL)
wasinjectedinthesplitlessmode(2minsplitlesstime)at350°C
usinga linerpackedwithglasswool Thecapillarycolumnwasa
15m× 0.32mm× 0.1μmDB5-HTcolumn(Agilent)andwas
op-erated at a constant flow of 2 mLmin−1 using helium as carrier
gas The temperatureprogram ran from60°C (3 min) to 350°C
(3min) at15°C/min.The FIDtemperaturewassetat350°C, hy-drogenandairflows were36and400 mLmin−1,respectivelyand the datacollection ratewas 200 Hz.The MOSH/MOAH composi-tionofthemineraloilswascalculatedusingthetotalareaandthe areasoftherespectivefractions
3 Results and discussion
3.1 SFC-FID/UV selectivity
Tounderstandthe interaction ofmineraloil constituentswith the stationary phase under SFC conditions, mixtures of standard compounds were injected first Fig 1 shows the SFC-FID chro-matograms obtained forthe separation ofthe aliphaticand aro-matic standards on the bare silica and the silver-loaded silica columnsusingpure carbondioxideasthemobilephase.All com-pounds elutein less than 20 min from both columns The alka-nes C11 and C13 coelute with the solvent peak (hexane) for the bare silica column (Fig 1a), indicating the absence of retention for these compounds Unfortunately, neither the bare silica col-umnnorthesilver-loadedsilicagiveafullseparationofthe com-poundsbasedonaromaticity.Unexpectedly,5-α-cholestane(Cho),
analiphaticcompoundthatwasexpectedtoshowlittleinteraction withthe stationaryphase, coeluteswiththearomatics.Moreover, forbothstationaryphases,1,4-di-dodecylbenzene(D12B),a mono-aromaticcompound,elutesafter1-methylnaphthalene(1MN)and biphenyl(BP),twodi-aromaticspecies.Aliphatic/aromatic selectiv-itywasbetterforthesilver-loadedsilicacolumn,whereChoeluted slightlybeforethedi-aromatic compounds.Thestrongerretention
ofthearomatics onthe silverloadedphase ismostlikely dueto
a stronginteraction of the aromaticπ electrons with the vacant orbitalsof the silverions [21,22] Nevertheless,based on the re-sultsofthisexperimentitisclearthatretentionandselectivityin SFCcannotbeexplainedbytheinteractionoftheanalyteswiththe stationaryphaseonly
To better understand the influence of the mobile phase on MOSH/MOAH retentionand selectivityin SFC, the samestandard compounds were also analyzed in NPLC using hexane as the mobile phase Fig 2 shows the chromatograms obtained for the NPLC-UV/RID and the SFC-FID separation of the aliphatic and aromaticstandardsonthe silver-loadedsilica column.In contrast
to the situation in SFC, there is no separation of the individual compounds in NPLC In the latter mode the analytes are clearly separated accordingto their aromaticity, with little effect of the size of the molecule or the presence of side chains [10] All aliphaticcompounds elutetogether atthebeginning ofthe chro-matogramfollowedbythemono-,di-andtri-aromaticcompounds
intheorderofincreasingaromaticity.Thisconfirmsthatretention andselectivityinSFCnotonlydependontheinteractionwiththe stationary phase but are also influenced by the solubility of the analytesintheCO2
Retention and selectivity in packed column SFC have been extensivelystudied inthe past [23] In thesestudies three main
Trang 4Fig 1 SFC-FID separation of aliphatic (black) and aromatic (red) standards on (a) bare silica and (b) silver-loaded silica columns Standards: C 11 (1), C 13 (2), Cycy (3), Cho (4), TBB (5), 1MN (6), BP (7), D12B (8) and Ant (9) CO 2 at 60 °C, 110 bar and 1 mLmin −1 was used as mobile phase (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
effects are identified: (i) solubility in the mobile phase, (ii)
in-teraction with the stationary phase-mobile phase complex (the
mobile phase, CO2, is absorbed on the stationary phase forming
amono-ormultilayer[24])and(iii)displacementoftheanalytes
fromthe stationary phase by the mobile phase molecules.Since
each ofthesemechanisms dependson temperatureandpressure
inits ownway, predictionof retentioninSFC isdifficultat best
Although the above mentioned three mechanisms-model is a
simplificationthatpartlyviolatestheprinciplethat in
chromatog-raphy mobile and stationary phase effects are inseparable, it is
helpfulinunderstandingthecurrentobservations.Apparently, for
thesystemstudiedhere,atthepressureandtemperatureapplied,
supercritical CO2 isa weaker solvent thanhexane Moreover,the
interaction ofthe analytes is not only with the stationaryphase
but with the stationary phase covered with CO2 As a result of
these effects, aliphatic compounds can be retained even though
theyare expectedto havelimitedinteraction withthe stationary
phase.Thus,retentionofthecompoundsinSFCisrelatedtotheir
aromaticityasthisdeterminestheinteraction withthestationary
phase,butalsobytheirmolecularweightwhichaffectstheir solu-bilityinthemobilephase.Asa result,smallaliphaticcompounds will elute rapidly, whereas larger aliphatic species will elute slowerduetotheirlowsolubility.Theycanthencoelutewiththe smallermono-aromaticcompounds.Inthesameway,large mono-aromaticcompounds can elute aftersmaller di-orpoly-aromatic compounds as is observed for 1,4-di-dodecylbenzene (D12B) However, if two molecules with a similar solubility, or similar molecular weight, have different aromaticity, they will be sepa-ratedsincetheir interactionwiththestationaryphaseisdifferent Whethera good MOSH/MOAH resolution isobtained dependson the difference in retention betweenthe ‘average MOSH’ and the
‘averageMOAH’ molecule ontheone hand,andthe widthofthe MOSHandMOAHbandsontheother.Thewidthofthesebandsis determined not just by chromatographic band broadening mech-anisms, but more importantly by retention differences between thevariousmoleculeswithinagroup,i.e.thepolydispersityofthe MOSH andMOAH fractions interms ofsize and structure ofthe molecules
Trang 5Fig 2 Chromatograms of the separation of the aliphatic (black) and aromatic (red) standard compounds on a silver-loaded silica column by (a) NPLC-UV/RID and (b) SFC-
FID Standards: C 11 (1), C 13 (2), Cycy (3), Cho (4), TBB (5), D12B (6), 1MN (7), BP (8) and Ant (9) Hexane was used as mobile phase in the NPLC-UV/RID separation CO 2 at
60 °C, 110 bar and 1 mLmin −1 was used as mobile phase in the SFC-FID separation (For interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
Fig 3 showsthe SFC-FID/UVMOSH/MOAH separation of
min-eral oil1on the silver-loadedsilica column Three peaksare
ob-servedinthechromatogram.Thefirstpeakisthesolvent,hexane,
followed by the MOSH and the MOAH bands Separation of the
MOSHandMOAHfractionsisconfirmedbytheUVchromatogram,
which evidently only shows the aromatic compounds The
res-olution obtained is insufficient for accurate MOSH and MOAH
quantification.Furtheroptimizationoftemperatureandpressureis
neededtoimproveMOSH/MOAHresolution
3.2 Optimization of the SFC-FID/UV method for mineral oil analysis
As previously described, retention in SFCis affected by three
mechanisms,thetwomostimportantonesbeingtheinteractionof
theanalytewiththeCO2-coveredstationaryphaseandthe
solubil-ityoftheanalytesintheCO2 mobilephase.Forthislatter
mecha-nism thedensityofthe CO2,i.e.whetherit ismoreliquid-likeor
moregas-like,isvery important.Highpressures andlow
temper-aturesgivetheCO2 amoreliquid-likenatureimprovingthe solu-bilityoftheanalytesandreducing retentionandbandbroadening duetothepolydispersityoftheMOSHandMOAHfractions.Onthe contrary,low pressures and hightemperatures decreasethe ana-lytesolubilityincreasingtheirretentionandpeakwidth[7] More-over, like in all forms of chromatography the mobile phase flow ratewillalso influenceanalyte retention,peak widthand resolu-tionoftheanalytes.Thus,optimizationofthemethodshould con-siderthesethreeparametersandidentifyconditionswhere maxi-mumresolutionisobtained
To identify the optimal chromatographic conditions for the MOSH/MOAHseparation,atwo-levelfactorialexperimentaldesign optimizationwasperformedwithtemperature, pressureandflow
as factors The high andlow levels chosen for the three factors areshowninTable2.Theresponseselected foroptimizationwas MOSH/MOAH resolution Resolution values obtained for the dif-ferent experiments are given in Table 3 For the MOSH fraction, peakwidthwasobtainedfromtheFIDchromatogram.MOAHpeak
Trang 6Fig 3 SFC-FID/UV separation of mineral oil 1 on a silver-loaded silica column First peak corresponds to the solvent, hexane, followed by MOSH and MOAH CO 2 at 60 °C,
110 bar and 1 mLmin −1 was used as mobile phase MOSH/MOAH resolution is 0.9057 The UV chromatogram was plotted using the MaxPlot mode
Table 3
Resolution values obtained for every experiment for the optimization of the SFC-FID/UV method Every combination was evaluated in triplicate
Experiment Temperature ( °C) Pressure (bar) Flow (mLmin −1 ) Resolution
1.0117 0.9883
1.0293 1.0151
1.0756 1.0324
0.9040 0.9447
0.9258 0.9464
1.0133 1.0245
1.0783 1.0595
1.0590 1.0603
width was obtained from the UV chromatogram where it could
be moreaccurately determined dueto the absenceof the MOSH
signal.Priortostartingtheevaluation,theFID andUV signalsfor
the MOAH band were compared to see if possibly extra-column
band-broadeningoccurredin theT-piece orthetransfer capillary
to the FID This was done using a pure MOAH fraction isolated
froman oil using SPE No extra band broadening was observed,
althougha slight shiftin theretention timeswas seen,with the
FID signal beingslightly later than that of the UV detector
Sta-tistical analysis using the Analysis of Variance (ANOVA) method
showedthat the flow rateapplied did not havea significant
in-fluenceonMOSH/MOAHresolution,whereastheeffectsof
temper-atureandpressureweresignificant.Detailsofthestatistical
analy-sis,theoptimization oftheresponse andtheequation describing
the influence of each parameter on MOSH/MOAH resolution are
shownin the supplementary material (SM-1) Fig 4shows reso-lution as a function of pressure and temperature at a flow rate
of1 mLmin−1 Resolution isseen to increase athigher pressures andtemperatures.Undertheseconditionsthealiphaticcompounds will have lessinteraction with thestationary phase and elutein
a narrowerband improvingtheresolutionbetweenthetwo min-eraloilfractions.Amaximumresolutionof1.08 isobtainedwhen workingat 138 bar and 60°C This resolution value is sufficient
if the MOSH and MOAH bands are of similar height, but is too low foran accurate MOSH/MOAH determinationof sampleswith high MOSH/MOAH ratios, such asencountered in highlypurified whiteoils.Giventheshapeoftheresolutionfunctionobservedin Fig 4, it waslogical to test even higherpressures and tempera-tures.Unfortunately,atthoseconditionstheMOSHbandstartedto overlap withthe solventpeak leadingto an incorrect
Trang 7quantifica-Fig 4 Response surface of the resolution as a function of temperature and pressure Flow is set at 1 mLmin −1 The maximum resolution, 1.08, is obtained at 60 °C and
138 bar
Fig 5 Deconvolution process of the MOSH and MOAH peaks in the FID chromatogram using the UV signal, (a) normalized SFC-FID and SFC-UV chromatograms, (b) SFC-FID
chromatogram after subtraction of the normalized UV signal, MOAH peak is removed, and (c) deconvoluted SFC-FID chromatogram of mineral oil 1 MOSH/MOAH separation was carried out on a silver-loaded silica column and CO 2 at 60 °C, 138 bar and 1 mLmin −1 was used as mobile phase The UV chromatogram was plotted using the MaxPlot mode
tion of the MOSH fraction Hence, the conditions chosen forthe
SFCMOSH/MOAHseparationwere138bar,60°Cand1mLmin−1
IncaseofdeterminationswhereonlytheMOAHlevelsareneeded,
workingathighertemperaturesandpressuresispossible.Pressure
gradientswerebrieflystudiedasameanstoimproveMOSH/MOAH
separation while avoiding coelution of the MOSH band withthe
solvent peak However, no large improvement wasobserved and
theoptionofpressureprogrammingwasnot furtherstudied.This
alsobecausequantificationwouldbecomplicatedasthesplitratio
betweenUVandFIDwillvaryduringpressureprogramming
Addi-tionally,theFIDresponsemightbeaffectedbytheincreasing CO2
flow
To solve the partial overlap issue, the possibility to
deconvo-lute theMOSHandMOAH peaksusingthecombinedFIDandUV
chromatogramswasstudied.The principleofthisapproachis
de-scribedinFig.5.Fig.5ashowstheoverlayoftheSFC-FIDand
SFC-UV chromatograms of mineral oil 1 at the optimal temperature
andpressureconditionsafternormalizationofthetwosignals.For theUV signal, the MaxPlot (190–400 nm)trace wasused asthis showedthebestsimilaritytotheFIDsignal.Usingthenormalized MOAHpeakfromtheSFC-UVchromatogram,theMOAH contribu-tionintheSFC-FIDchromatogram canberemovedbysubtraction (seeFig.5b)leavingjustthesignaloftheMOSH.Normalizationof theresponsewasdoneusingthefollowingequation:
FIDnew(t)=
UV(t) ( FID)
UVmax
+FIDmin where FIDnew(t) is the MOAH FID signal at time t, UV(t) repre-sentstheUVsignal attimet, FID isthedifferencebetweenthe FID response at the time of the maximum UV response and the baselineFIDresponsealongthechromatogram,UVmaxisthe max-imumUVresponsealongtheUV chromatogramandFIDmin isthe minimum,i.e.baseline, FID response.Fig 5cshowsthe deconvo-lutedSFC-FID chromatogramof mineraloil 1 andits comparison
Trang 8Fig 6 Comparison of the deconvoluted SFC-FID chromatogram of mineral oil 1 and the SFC-FID chromatograms of the individual MOSH and MOAH fractions of the same
mineral oil after separation by SPE
with the original chromatogram Both the individual MOSH and
MOAHsignalsmatchperfectlywiththeoriginalchromatogram,but
thetwo signals are now separated.With the deconvoluted
chro-matogramsbetterintegration andmoreaccuratequantification of
thetwofractionsis possible.Thecomparisonofquantitative data
obtainedusingthisdeconvolutionstrategywiththoseobtainedby
thestandardSPE-GC-FIDmethodwillbediscussedinthenext
sec-tions The FID MOSH/MOAH deconvolution was further
corrobo-rated by injecting isolated MOSH and MOAH fractions, obtained
usingSPE,individuallyintotheSFCsystem.Fig.6showsthe
com-parison of the deconvoluted FID chromatogram and the SFC-FID
chromatogramsof the individual MOSH andMOAH fractions
Al-thoughthedeconvolutedbandsandtheSPEfractionsdonotmatch
perfectlydueto smalldifferencesinconcentrationsandretention
timeshifts,thefigureclearlyshowsthattheMOSH/MOAH
decon-volutiongivesverysimilarpeakstothoseobtainedforthe
individ-ualfractions.Thisconfirmsthat theUV signaloftheSFCanalysis
canbeusedfordeconvolutionoftheFIDsignalinordertoallowa
moreaccurate integrationoftheMOSHandMOAHpeaksand
ob-tainmoreaccuratequantificationofthetwofractions
3.3 Validation of the SFC-FID/UV method
In order to evaluate the performance of the proposed
SFC-FID/UVmethod,themainperformancecharacteristics,i.e.linearity,
repeatability, limit of detection (LOD) and limit of quantification
(LOQ)wereassessedfollowingtheEurachemguideformethod
val-idation[20] Thisevaluationwasdone usingmineraloil 8,an oil
free of aromatics,hence only the MOSH fraction wasconsidered
toassesstheperformancecharacteristics.Trueness ofthe method
wasassessed by comparing the results ofthe new method with
that of off-line SPE fractionation and GC-FID quantification (see
Section3.4) SincetheFID respondsunselectivelytohydrocarbons
andMOSHandMOAHhaveverysimilarresponsefactors,the
per-formancecharacteristicsdeterminedusingMOSHarealsovalidfor
MOAH[1].Theresultsobtainedinthevalidationexperimentsand
thecomparison with the performance characteristics of the
SPE-GC-FIDmethodarepresentedinTable4andthecalculationsand
graphs are shownin the supplementary material (SM-2) An
ex-Table 4
Performance characteristics for the validation of the SFC-FID/UV and SPE-GC-FID methods
Linearity 0.9995 0.9641 LOD (mgmL −1 ) 0.0360 0.0191
LOQ (mgmL −1 ) 0.1200 0.0637
Repeatability (CV%) < 4.84 < 6.77
cellent correlation (R2 > 0.9995) between the FID response and themineraloilconcentrationwasfoundintheconcentrationrange 2.5–100mgmL−1.Repeatability ofthemethod,evaluated atthree concentration levels, resulted in RSD values below 5% Unlike in classicaltwo-stepMOSH/MOAHmethods,variationoftheresultsis onlyrelatedtothepeakintegrationsincenoothersamplehandling stepsarepresent LODandLOQ,evaluated statisticallyastheRSD
of the injection of three different samples at low concentration, were0.0360and0.1200mgmL−1,respectively.Thus,whenworking withamineraloilconcentration of10mgmL−1 andconsideringa mineraloilwithsimilarphysicochemicalcharacteristicsasmineral oil 8, theLOD andLOQ values forthe MOAH fraction calculated
asthe percentageof MOAH inthe sampleare 0.36and 1.2%, re-spectively.This issufficientlysensitive forall butthe highest pu-rity white oils It should be noted here that the LOQ cannot be improvedby injectinga largersamplevolumeora more concen-tratedsample.ThisbecauseLOQ/LODvaluesarenotdeterminedby thedetectorsensitivity,butratherbytheMOSH/MOAHseparation, with thisseparation being compromised when the MOSH/MOAH ratioofthe samplegets toohigh Toconfirm thetrueness ofthe calculatedLOQvalueforMOAH,asolutionat10mgmL−1ofa mix-tureof mineraloil1 (24%aromatics)andmineraloil 8(virtually aromaticfree), witha final MOAH concentration of1.2%, was in-jected in the SFC-FID/UV system After deconvolution of the FID chromatogram andcalculationof thearomatics contentbased on the MOSH/MOAH area ratios, an aromatics level of 2% was ob-tained, so well above the calculated value of 1.2% Fig 7 shows
Trang 9Fig 7 SFC-UV chromatograms of mineral oil 8 and a mixture of mineral oil 1 and 8 with a final MOAH concentration of 1.2% with respect to the total mineral oil The
double MOAH peak in the mixture could be the result of a high naphthene concentration in mineral oil 8 CO 2 at 60 °C, 138 bar and 1 mLmin −1 was used as mobile phase
UV chromatograms were plotted using the MaxPlot mode
plotted using the MaxPlot mode
theSFC-UVchromatogramsofmineraloil8andofthemixtureof
mineraloil1and8previouslydescribed.Unexpectedly,theSFC-UV
MaxPlot chromatogram ofthearomatics-free mineraloil8 shows
apeak around3minwhichmatches withthefirstpeakobserved
in the chromatogram of the mixture This elution position
sug-geststhesignalisnotcausedbyaromaticspeciesbutresultsfrom
aliphatics.Knowingthatinthepreparationofaromatics-free
min-eral oils hydrogenationis applied andthat thiswill resultinthe
formationofnaphthenes,i.e.cyclicaliphatics,theSFC-UVresponse
of the two naphthenes in the set of standard compounds, Cycy
andCho,wasstudied Fig.8showstheresulting SFC-UVMaxPlot
chromatogram Even though no response of the two naphthenes
wasexpected,twoclearpeaksareobservedinthechromatogram Comparisonoftheabsorbancespectrumofthestandardsandthe peak seenfor mineraloil8 showedvery similar spectra that are clearly different from those obtained for aromatic species As a consequence,the UVMaxPlot signalofthemixtureofmineraloil
1and8isformedbytwohumps,oneforthenaphthenesof min-eral oil8 eluting in the MOSH window, and one for the MOAH PlottingtheUV chromatogramusingthe aromaticswavelengthof
254 nm rather than the MaxPlot mode confirms that the peak around3minarenotaromaticspecies.Apparentlythenaphthenes show astrong absorbanceintheMaxPlot region of190–400 nm Deconvolutionof theFID signalconsidering onlythe MOAHpeak
Trang 10Fig 9 SFC-FID chromatograms of mineral oil 1 (black), mineral oil 5 (red) and mineral oil 6 (blue) The three mineral oils have different viscosities of 16.53, 56.41 and
91.21 mm 2 s −1 , respectively MOAH peak is visible in mineral oil 1 and 5 but not in mineral oil 6 (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig 10 Comparison of the aromatic percentage of five different mineral oil samples using the SFC-FID/UV and the off-line SPE-GC-FID standard method
resultedinanaromaticcontentof1.4% inthemixtureofmineral
oils,avalueclosetotheexpectedlevel,corroboratingthetrueness
ofthemethod
3.4 Analysis of real samples
Following the optimization and validation of the SFC-FID/UV
method,seven different mineral oils fromdifferent stages inthe
whiteoilpurificationprocesswereanalyzed.Themineraloilswere
chosen with the intention to cover a wide range of samples in
termsofviscositiesandmolecular weights.Fig 9showsthe
SFC-FID chromatograms ofthree different mineraloils (mineral oil 1,
5 and 6) Mineral oil 6 has the broadest MOSH hump followed
by mineraloil5and1.The highertheviscosityorthe molecular weightofthesample,thebroadertheMOSHandMOAHhumps.A broaderMOSHhumpresultsinmoreseverecoelutionoftheMOSH and MOAH humps complicating the deconvolution process The MOAHhumpisvisibleformineraloils1and5,butnotformineral oil6.Consequently,theSFC-FID/UVmethodforMOSH/MOAH anal-ysisislimitedtolightermineraloilswithaviscosityandmolecular weightbelowapproximately56mm2 −1 and450gmol−1,
respec-tively Fig 10 compares the MOAH composition of five different