Practical observations on the performance of bare silica in hydrophilic interaction compared with C18 reversed phase liquid chromatography

ContentslistsavailableatScienceDirect

Journal

of

Chromatography

A

j o ur na l h o me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / c h r o m a

Practical

observations

on

the

performance

of

bare

silica

in

hydrophilic

interaction

compared

with

C18

reversed-phase

liquid

chromatography

夽

James

C.

Heaton

,

Xiaoli

Wang

,

William

E.

Barber

,

Stephan

M.C.

Buckenmaier

,

David

V.

McCalley

a_{CentreforResearchinBiosciences,UniversityoftheWestofEngland,Frenchay,BristolBS161QY,UK} b_{AgilentTechnologiesInc.,2850CentervilleRoad,Wilmington,DE19808,USA}

c_{AgilentTechnologies,Hewlett-PackardStraße8,76337Waldbronn,Germany}

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received2September2013 Receivedinrevisedform 17December2013 Accepted18December2013

Available online 27 December 2013

Keywords:

HILIC

Retentionmechanism Peak-parking

b-Term

c-Term

a

b

s

t

r

a

c

t

ThekineticperformanceofabaresilicaandC18phasepreparedfromthesamesub-2␮mand3.5␮mbase materialswerecomparedintheHILICandRPmodeusingbothchargedandneutralsolutes.TheHILIC columnwascharacterisedusingtheneutralsolute5-hydroxymethyluridine,theweakbasecytosine, andthestrongbasenortriptyline,thelatterhavingsufficientretentionalsointheRPmodetoallow comparisonofperformance.NaphthalenewasalsousedasasimpleneutralsubstancetoevaluatetheRP columnalone.Theretentionfactorsofallsubstanceswereadjustedtogivesimilarvalues(k∼5.5)attheir respectiveoptimumlinearvelocities.ReducedvanDeemterb-coefficients(determinedbycurvefitting andbythepeakparkingmethod,usinganovelprocedureinvolvingswitchingtoadummycolumn)were significantlylowerinHILICforallsubstancescomparedwiththosefoundunderRPconditions.Against expectation,c-coefficientswerealwayslowerinRPwhencomparedwithHILICusingsub-2␮mparticles. Whilemeasurementofthesecoefficientsiscomplicatedbyretentionshiftscausedbytheinfluenceof highpressureandbyfrictionalheatingeffects,broadlysimilarresultswereobtainedonlargerparticle (3.5␮m)phases.Themechanismoftheseparationswasfurtherinvestigatedbyexaminingtheeffectof bufferconcentrationonretention.ItwasconcludedthatHILICcansometimesshowsomewhatinferior performancetoRPforfastanalysisathighmobilephasevelocity,butclearlyshowsadvantageswhen highcolumnefficiencies,usinglongercolumnsatlowflowvelocity,areemployed.Thelatterresultis attributabletothelowerviscosityofthemobilephaseinHILICandthereducedpressurerequirementas wellasthelowerb-coefficients.

© 2014 David V. McCalley. Published by Elsevier B.V.

1. Introduction

Hydrophilicinteractionchromatography(HILIC)hasbeen gain-ingrapidacceptancefortheanalysisofhydrophilic/polar/charged solutessincetheearlyworkofAlpertetal.[1]in1990.Arguably however,itwasMartinandSynge[2]whofirstemployeda HILIC-typeseparation,usingawater-saturatedbaresilicacolumn,albeit withthewater-immisciblesolventchloroformasthemobilephase, toseparatepolaraminoacids.Theretentionmechanismwaslikely to be similar in both studies. In HILIC, partitioning, hydrogen

夽 _{Presentedatthe39thInternationalSymposiumonHigh-Performance}

Liquid-PhaseSeparationsandRelatedTechniques,Amsterdam,Netherlands,16–20June 2013.

∗Correspondingauthor.Tel.:+441173282469;fax:+441173282904.

E-mailaddress:david.mccalley@uwe.ac.uk(D.V.McCalley).

bonding and coulombicinteractions are consideredtopromote retention,usuallyinacetonitrilerich(>70%,v/v)bufferedeluents onpolarstationaryphases.Thepresenceofawater-richlayeratthe silanolinterfaceonsilicaphaseshasbeenconfirmedbyboth exper-iment[3,4]andviamoleculardynamicsimulationapproaches[5,6]. Itisbetweenthislayerandthebulkacetonitrile-richmobilephase thatanalytepartitioningisthoughttotakeplace.However, depend-ingonthenatureoftheanalyte,partitioningonlycontributesin somemeasuretotheoverallretentionmechanism.Forinstance, McCalley[7]determinedthatalargeproportionoftheretention forsomebasicsolutesonbaresilicawasgovernedbyionic pro-cesses.Therearenowmanycommerciallyavailablepolarbonded phases which can be used in HILIC providing the user with a plethora ofmethoddevelopmentchoices.Irgum[8],Tanaka[9]

and McCalley[10] andtheirco-workershave madeprogressin attemptingtocharacterisetheselectivityof thesephasesunder various operating conditions. Unlike reversed-phase (RP),there

0021-9673© 2014 David V. McCalley. Published by Elsevier B.V.

http://dx.doi.org/10.1016/j.chroma.2013.12.058

Open access under CC BY license.

remainsdifficultyinaddressinggeneralapproaches to develop-ingmethodsinHILIC,despitethemanyadvantagesitpossesses, whichisaconsequenceoftheill-definedseparationmechanism. Inadditiontotheenhancedretentionofpolarcompounds,there areseveraldistinctadvantagesofHILICoverRP.Theseincludethe volatilityofACN-richmobilephases(easeofcouplingandbetter sensitivitywithmassspectrometry[11]),lowoperatingpressures toachieve a given linear velocity [12], and different retention selectivitycomparedwithRP[13],offeringforexamplethe poten-tialfor two-dimensionalseparations.Superiorpeakshapesmay alsobeobtainedforsomecompoundsinHILICcomparedwithRP

[14,15].

Despitethesenumerousstudies,relativelylittleworkhasbeen publishedconcerningthekineticperformanceofHILICcompared withRP.Simple vanDeemterplotsshowthat sometimes, supe-rior mass transfer is obtained in HILIC for basic compounds, duetothelowerviscosityofthemobilephasesusedand resul-tant enhanced solute diffusivity, and that also longer columns wereapplicablegeneratinghighnumbersoftheoreticalplatesin areasonable analysistime [12].However,reduced plots, which would allow a more fundamental comparison of these tech-niques,werenotpresentedinthatstudy.Inthecurrentwork,a moredetailedcomparisonofthekineticperformanceofthe tech-niqueswasattempted.We usedavariety oftest soluteswhich coverabreadthofphysiochemicalpropertiesinordertoaddress the practical comparison of HILIC versus RP. These substances includedahydrophilic-neutralsolute(5-(hydroxymethyl)uridine), hydrophilic-weakbase(cytosine),hydrophobic-strongbase (nor-triptyline)andahydrophobic-neutral(naphthalene).Thisrangeof compoundswasemployedasHILICisoftenappliedtoionisedas wellaspolarneutralspecies.Acidswerenotincludedinthestudy astheyoftengiveverypoorretentiononbaresilicaHILICphases duetorepulsionfromionisedsilanolgroups[10].Weemployed thesamenativebasesilicafromthesamemanufacturerforboth RPandHILICinallcasessothatabettercomparisonofthe tech-niquescouldbeobtained. Theb-coefficientswerederived from peakparkingexperimentsusinganovelvalvesystemusingan addi-tionalreferencecolumntoallowimmediatepressurisationofthe measurementcolumnaftertheparkingperiod.Bulkdiffusion coef-ficients(Dm)weremeasuredexperimentallyusingtheTaylor–Aris procedure[12]andcomparedwithvaluesobtainedfromestimates usingtheWilke–Changandotherprocedures[16].Peakparking derivedb-coefficientswerethencomparedwiththosefromsimple curvefittingofplotsofreducedplateheightagainstflow. Measure-mentsweremadewithmatchedsub2-␮mand3.5␮mphases,in ordertodetermineanyeffectofhigherpressureonthe measure-ments.Finally,theeffectofbufferconcentrationonretentionwas investigatedinordertoinformtheinterpretationoftheretention mechanismintheseexperiments.

2. Experimental

2.1. Chemicalsandreagents

HPLCgrade acetonitrilewaspurchasedfromFisherScientific (Loughborough,U.K.).Ammoniumformate,formicacid, toluene, naphthalene,uracil,5-(hydroxymethyl)uridine,cytosineand nor-triptylinewere all purchased fromSigma–Aldrich (Poole, U.K.). Waterat18.2mwasfromaPuriteOnedeopurifier(Thame,UK).

2.2. Mobilephasepreparation

Pre-mixedeluents were preparedby weighing the solvents, convertingtothe %(v/v)values shownin Table1 bymeansof theirdensity.Stockbuffersolutionsof100mMand125mMwere

Table1

Preparationofmobilephasesusedinthepeakparkingandflowstudies.

Analyte Mode %ACN Bufferconcentration (mM/L)

Cytosine HILIC 92.7 5.24

Nortriptyline HILIC 94.2 5.25 5-(Hydroxymethyl)uridine HILIC 94.8 5.93

Naphthalene RP 54.9 n/a

Nortriptyline RP 33.4 20.3

prepared by dissolving the appropriate amount of ammonium formateinwaterandadjustingtopH3usingformicacid.

2.3. Apparatusandmethodology

Table2

Non-reducedvanDeemtercoefficientsfromcurvefitting.

Analyte Mode A(×10−4_{cm) B(×10}−5_cm2_{/s) C(ms) u}

opt(cm/s) Hmin(×10−4cm) Nmax

dp=1.8␮m

Cytosine HILIC 2.2 3.7 0.37 0.32 4.7 10,700

5-(OHmethyl)uridine HILIC 1.9 3.2 0.61 0.23 4.7 10,300

Nortriptyline HILIC 2.3 4.2 0.34 0.35 4.9 10,700

Nortriptyline RP 2.5 2.8 0.47 0.24 4.8 10,400

Naphthalene RP 2.0 8.4 0.18 0.69 4.3 11,500

dp=3.5␮m

Cytosine HILIC 4.5 3.8 0.98 0.20 8.3 6050

5-(OHmethyl)uridine HILIC 4.7 2.9 1.42 0.14 8.4 5990

Nortriptyline HILIC 4.6 4.4 0.82 0.23 8.3 5950

Nortriptyline RP 4.5 2.9 1.17 0.16 8.0 6270

Naphthalene RP 3.5 8.7 0.54 0.40 7.9 6310

methodusingaflowrateof0.1mL/minat30◦C[12].Theinternal diameterandlengthofthePEEKtubingusedwas0.05277cmand 1500cm,andthecoildiameterwas22cm.

3. Resultsanddiscussion

3.1. SimplevanDeemterplots

InbothHILICandRP,themobilephasestrengthwasadjusted to give similar k (∼5.5) near their respective optimum linear velocitiesinordertominimisethecorrectionsforextra-column bandspreading.Thesecorrectionswereonlyoftheorderofafew %.ThedatawasfittedtothevanDeemterequation:

H=A+B_u+Cu (1)

whereHistheheightequivalenttoatheoreticalplate,uisthe lin-earvelocity,andA,BandCareconstantsthatcanbederivedfrom curvefittingofHversusudata.Throughoutthis workwerefer tonon-reducedcoefficientsinuppercase,whilereduced coeffi-cientsareinlowercase.Wecomparedpairsofboth1.8␮mand 3.5␮mparticlepackedcolumns.Fig.1(a)and(c)(plotsofHversus u)showtheexpectedsmallerplateheightsforthesmaller parti-cle1.8␮mHILICandreversed-phasecolumns.Weusedaslightly higherbufferconcentrationfor5-(hydroxymethyluridine)inHILIC inordertomaintainkat5.5withoutusinglessthan5%waterin themobilephase,whichisacommonlowerlimitemployed.Note theretentionofthissoluteincreaseswithincreasingbuffer con-centration(seebelow).Wealsousedhigherbufferconcentrations (∼20mM)forthenortriptylinereversed-phaseworksoasto min-imiseoverloadingeffectsonthepeakshape,whichmaybeapparent evenatverylowsampleconcentrations.Forthisreasonwedidnot generallyemploythestatisticalmomentsapproachforcalculating peakprofilesduetothelowS/N(<100)valuesencountered,which canunderminetheprecisionandaccuracyofsuchmeasurements. TheUSPtailingfactors(5%ofpeakheight)fornortriptylineonthe 1.8␮mand3.5␮mphasesweregood (worstvalues1.2 and1.1 respectively),measuredattheirrespectiveoptimumlinear veloci-ties.IndeedFig.1(e)and(f)demonstratethatforthesesymmetric peaks,littledifferencewasshowninthecurvesfornortriptyline inHILICornaphthaleneinRPwhentheplateheightwas calcu-latedbythehalf-height,5sigma(4.4%peakheight)orstatistical momentsmethod(dataisshownforthesmallerparticlecolumns). Resultswouldbeexpectedtodiffermoresubstantiallyinthecase ofasymmetricsolutepeaks.

Table2summarisesthevanDeemterA,BandCcoefficientsfor thetwoevaluatedparticlesizes.Verysimilarresultswereobtained onboththe1.8␮mand3.5␮mcolumns.Itcanbeseenonboththat theB-coefficientfornortriptyline(RP)issmallerthanthatforthe samesoluteunderHILICconditions.Furthermore,theC-coefficient fornortriptyline(RP)islargerthanthatforthesamesoluteunder

HILICconditions.Thesefindingsagreewithpreviousresultsfor nor-triptyline[12]andareinlinewithexpectationconcerningHILIC versusRP.Table3showsthattheexperimentallymeasuredbulk diffusioncoefficient(Dm)ofnortriptylineisconsiderablygreater (morethantwice)intheHILICeluent(94.2%ACN)thanthatused for RP (33.4%), which as expected leads to greater axial diffu-sion(increasedB-coefficient)butimprovedmasstransfer(reduced C-coefficient)inHILIC.Nevertheless,whereasnaphthalene(RP), 5-(hydroxymethyluridine(HILIC)andnortriptyline(HILIC)havevery similarDmintheirrespectivemobilephases,theB-coefficientfor naphthalenewasconsiderablylarger,andtheC-coefficientsmaller thanforboththeHILICsolutes.Thisresultisunexpected,and sug-geststhatotherfactorsthantheincreasedbulkdiffusivityinHILIC mobilephasesareinvolved.

Itispossiblethatfrictionalheatingeffectsmightcontributeto theC-coefficientsathighflowrate.HILICmobilephaseshavelower thermalconductivitywhichcouldcontributetoradialtemperature gradients.However,thelowviscosityofHILICmobilephaseand consequentloweroperatingpressuresatthesamelinearvelocity leadstolesspowergenerationandthusmayevenouttheeffects

[18].Similarresultswereshown forbothparticlesizecolumns, whichmayindicateindeedthatfrictionalheating(whichshould beconsiderablyworseforthe1.8␮mcolumns)had notgrossly affectedtheresults.

3.2. Determinationofb-coefficientsusingthearrestedelution method

Inanattempttointerpretfurthertheinitialobservationswe measuredthe(reduced)b-coefficientusingthearrested elution method[19,20].Briefly,thisprocedureentailsmeasuringthe effec-tivediffusionthroughthepackedbed(Deff)bystoppingtheeluent flowoncethesolutebandhasmigratedabouthalfwaydownthe column.Theflowisthen resumedafterasetperiodoftime (in whichdiffusionofthesolutetakesplace)andthebandiselutedout ofthecolumntothedetector.Thepeakspatialvarianceisrelated totheparkingtime(tpark)throughtherelationship:

2

x=2Deff·tpark (2)

Theequationcanbewrittenintermsoftemporalvarianceas:

Deff=t2·u2R/2tpark (3)

whereuRistheretainedpeakvelocityL/tRandtRisthetimeneeded topassthroughthecolumnwithlengthLwithoutstoppingtheflow. Deffcanthenbedeterminedfromaplotof_t2againstparkingtime. Knowledgeoftheanalytediffusioncoefficientinthebulkmobile phase(Dm)thenallowsforaccuratecalculationofthereduced b-coefficientfromEq.(4)[19].

b=2Deff Dm(1+k

Fig.1. Non-reducedandreducedvanDeemterplotsfor1.8␮mcolumns(a)and(b)andfor3.5␮mcolumns(c)and(d)respectively,usingthehalfheightmethodto determineplateheight.Non-reducedplotsfornortriptylineinHILIC(e)andnaphthaleneinRPmode(f)comparingthehalfheight,5-sigmaandstatisticalmomentsmethod fordeterminationofplateheight.Thek_{foreachanalyteneartheoptimumflowvelocitywasaround5.5.MobilephaseasshowninTable1.Somedatapointstakenatlower}

valuesofuandareomittedtoimprovetheclarityoffiguresa–d.

Carr[16]showedthatcorrelationsderivedfromempirical equa-tionsfordeterminationofDm(suchastheWilke–Changequation) broke down in acetonitrile rich mobile phases. We therefore determined diffusion coefficients experimentally, checking our

Table3

Comparisonofcalculatedversusexperimentaldiffusioncoefficients.SolventviscositywasdeterminedaccordingtoRef.[43].

T=30◦C

Solute %ACN (cP) Calculated Experimental

Dm (Scheibel) Dm (Wilke–Chang) Dm (Li–Carr) Dm (Taylor–Aris) %RSD(n=3)

Cytosine 92.7 0.3968 3.29E−05 3.02E−05 3.33E−05 1.77E−05 0.40 Nortriptyline 94.2 0.3823 1.50E−05 1.49E−05 1.75E−05 1.43E−05 0.43 Nortriptyline 33.4 0.7892 6.42E−06 7.49E−06 6.07E−06 5.85E−06 0.56 Naphthalene 54.9 0.6818 1.14E−05 1.33E−05 1.19E−05 1.38E−05 0.18 5-(Hydroxymethyl)uridine 94.8 0.3777 1.89E−05 1.90E−05 2.25E−05 1.42E−05 0.70

Thiourea(25◦_{C) 0 1.32E−05 0.92}

Dmvaluesincm2/s

Table4

EffectivediffusiondataandcomparisonofreducedvanDeemtercoefficients.

Analyte Mode Deff(cm2/s) Deff/Dm b(PeakParking) ba(Linearisation) b(CurveFitting) c(CurveFitting) opt hmin

dp=1.8␮m

Cytosine HILIC 2.58E−06 0.15 2.11 2.12 2.10 0.17 3.6 2.4

5-(OHmethyl)uridine HILIC 1.93E−06 0.14 1.94 2.27 2.26 0.22 3.2 2.4 Nortriptyline HILIC 3.09E−06 0.22 3.06 3.04 2.94 0.12 4.9 2.5 Nortriptyline RP 1.93E−06 0.33 4.97 4.67 4.74 0.07 8.2 2.4

Naphthalene RP 5.26E−06 0.38 5.73 6.00 6.06 0.06 9.8 2.2

dp=3.5␮m

Cytosine HILIC 3.01E−06 0.17 2.33 2.06 2.12 0.14 3.9 2.4

5-(OHmethyl)uridine HILIC 2.10E−06 0.15 2.21 1.92 2.01 0.17 3.5 2.4 Nortriptyline HILIC 3.61E−06 0.25 3.35 3.06 3.08 0.10 5.7 2.4 Nortriptyline RP 2.15E−06 0.37 5.46 4.91 4.99 0.06 9.4 2.3 Naphthalene RP 6.12E−06 0.44 6.47 6.28 6.31 0.06 10.1 2.3

a_{Linearisationmethod(interceptofhversus1/usingn=5datapoints,r}2_{>0.99inallcases).}

experimentallyforsomesolutesunderHILICconditions.For exam-ple,averylargedifferencewasfoundbetweentheWilke–Chang derived (3.02E−05cm2_{/s) and the experimentally determined} (1.77E−05cm2_{/s)valuesforcytosinein92.7%ACN.Thissignifies} thatseriouserrorswouldbeencounteredindeterminingreduced vanDeemtercoefficientsfortheb-andc-termsusingcalculated Dm values.Fig.2 showsplotsofpeakvarianceversusstoptime forthesoluteset(HILIC andRP)for bothparticlesizecolumns. Excellentlinearitywasobservedforallexperimentswith>0.999 r2_{valuesobtainedforeachsolute.Fig.1(b)and(d)showthevan} Deemterdataplottedinreducedcoordinateswherethereduced plateheight,h=H/dpandthereducedvelocity=udp/Dm.TheHILIC andRPcolumnsofnominally1.8and3.5␮mparticlesizeshowed verysimilarreducedplateheightsforallsolutesimplyingtheywere ofsimilarpackingquality.Simplifiedtheoryindicatesthatplotsof hversusfor“good”columnsoperatedunderthesameconditions shouldoverlay,and beindependentofdp[23].Typically,totally porousparticlecolumnsgivevaluesofhminbetween2.0and2.5, althoughthisisanempiricalobservation;suchvaluesareindeed showninTable4.Wedeliberatelymeasuredcurvesforanalytesat nearequivalenthighk,whereextra-columneffectsonlyminimally influencetheresults.

Table4showsvaluesofD_effcalculatedfromthepeakparking method,andvaluesofthereducedb-coefficientcalculatedfrom Eq.(4).Valuesofb-canbecomparedwiththosefromcurve fit-tingofthedatainFig.1(b)and(d).Linearisationanalysis[24]was alsousedtocross-checkb-coefficientsfromcurvefitting,asa suf-ficientnumberofdatapointswereobtainedinthelowflowregion toperformthisanalysis.Verygoodagreementisshownbetween thecurvefittingandpeakparkingmeasurementsforallsolutes andconditions.It mightbeconsideredthatsmalldifferencesin b-coefficientsmeasuredbyeithermethodareduetoshiftsink atdifferent flow rates inthe curvefittingmethod (seebelow), becausepeakparkingexperimentsinvolveuseofthesameflow ratetomovethepeakthroughandoutofthecolumn.However, thereislittledifferenceinthecomparedsetsofvaluesforthe1.8 and3.5␮mcolumns.Anadvantageofourpeakparkingprocedure

wasthatthearrestedcolumnwasheldunderpressurewhileflow wasdivertedtoadummycolumnusingaswitchingvalve.Once theflowisresumedonthecolumncontainingthepeak,minimal pressureperturbationisexperienced.Itcouldbeadvantageousto performarrestedelutionexperimentsinthiswayasithasbeen arguedelsewhere[25]thatitisimportanttodampentheeffectof pressureperturbationsduetostoppingandstartingflow.

Table4clearlyconfirmsthetentativeconclusionofsmaller b-coefficientsinHILICcomparedwithRP.Forexample,inreduced co-ordinates,theb-coefficientofnortriptylineinHILICis consid-erablysmallerthanthatforthesamesoluteinRPmode.Indeed, theb-coefficientsforbothnortriptylineandnaphthalene(RP)were greaterthanforanyoftheHILICsolutes.Thesmallerb-coefficients resultfromthesmallervaluesofDeff/DmforallthesolutesinHILIC comparedwiththoseinRP.Theobservationofasmallerb-termsin HILICversusRPwasveryrecentlydiscussedbyGrittiandGuiochon

[26].Indeed, theb-coefficientvalues inourstudyare in agree-mentwiththesepreviousobservationswhichwereoftheorder of2–3forHILICandbetween5–7forwellretainedsolutesinRP. InRPchromatography,surfacediffusionprovidesaconsiderable contributiontotheb-termofnon-polarsolutesretainedbya non-localisedretention mechanism.Theb-termmaybemade upof threedistinctcontributions[27].Firstly,diffusionoccursoutside theparticleinthemobilephase,secondlyinthemesoporesofthe particleandthirdlyalongtheporesurfaceofthestationaryphase (surfacediffusion)[28].ForhydrophobicneutralsubstancesinRP, surfacediffusionisfacilitatedbytheorganic-richlayerassociated withtheC18ligands.The surfacediffusioneffectisunderstood toaccount fora largeproportionof the intra-particlediffusion characterofaretainedsoluteinRP[29].Apparently,the mobil-ityofanalytesinteractingwithasolidsilicasurfaceintheHILIC modeismorerestricted.Ithasbeensuggestedthatthisimpliesan adsorptionmechanisminHILIC(atleastwithbaresilicacolumns) inwhichlocalisedinteractionsofthesolutetakeplacewithspecific silanolsitesofvariouskindsonthephasesurface[30]. Addition-ally,althoughsurfacediffusioncouldalsobeconsideredtooccur inHILIC,itwouldofnecessityhavetotakeplaceinahigh viscos-itywaterlayerandthuswouldcontributelittletotheintra-particle diffusivity.Theviscosityofpurewateris2–3timesgreaterthanthat ofpureacetonitrile[31].Assolutediffusivityisinverselyrelatedto theviscosity,surfacediffusionintheHILICwaterlayerwouldbe expectedtobeconsiderablylessthanthatinthesurfacelayerof ACNassociatedwithaC18stationaryphase.Localisationhasbeen givenasthereasonforsmallb-coefficientsmeasuredinadsorption chromatography(normalphase)withbaresilicacolumnsin stud-iesperformedmorethan30yearsagobySnyderandco-workers

[32].Indeed,adsorptionisacomponentoftheretention mecha-nisminHILICthatislikelytobeencouragedonbaresilicacolumns which haveless extensive water layers[4],especially withthe ratherlowmobilephasewaterconcentrationsusedinthepresent experiments[3].

Wehaveperformedthesestudieswithasinglemobilephaseper sampleandcolumngivinghighk.Previousstudieshaveshownthat inRPthereducedb-coefficientreachesabroadmaximumvalueat highk,whereasstabilisationofboccursrapidlyinHILIC[26].We believethatthisapproachallowsamorereliablecomparisonofthe modestobemade.

Whilereducedplotsaremoreusefulininterpretingunderlying processes,considerationofnon-reducedplotsremainsof impor-tancetothepractitioner,astheseplotsdemonstratewhatwillbe observedbyanexperimentalist(seealsokineticplotsbelow).

3.3. Comparisonofc-coefficients

ItisclearthatfromthereducedplotsinFig.1thatanalysisat highvelocity givesrisetolargerreducedplateheights inHILIC

comparedwithRP,whilehvaluesattheoptimumflowvelocity areremarkablysimilar.Itispossibletoexplainthisresultasbeing duetosloweradsorption–desorptionkineticsinHILIC,although there is littlefirm evidence in theliterature tosupport sucha hypothesis. Alternatively,Guiochon [30]hassuggested thatthe majorityofthisbandbroadeningeffectisduetolongrangeeddy dispersion.Itwasarguedthatthetransversediffusioncoefficient (whichisscaledtothediffusioncoefficientoftheanalyte)istoo smallandthecolumnIDtoolargetoallowcompleterelaxationof theradialconcentrationgradientwithinthecolumn. Thiseffect wouldthereforebegreaterinreducedplotsforHILICcompared withRP,duetolowersurfacediffusion.Gradientsmayoriginate at the column inlet due toa non-uniform sample distribution andatthecolumnoutletduetoasynchronoussamplecollection whenstandardendfittingsareused[30].Itwasnotedthatsome variationinthetransversediffusioncoefficientsmightbeobtained from one average particle diameter to another, which could explainthedifferencesinthec-termregionofthereducedplotsin

Fig.1(b)and(d).Theeffectswouldbereducedwithlongnarrow columns, providing that the bed structure of narrow columns showedgoodradialhomogeneityofthebedstructure(whichof course,is not oftenthe case).Thus at leastfor columns ofthe samedimensions,fastanalysisshouldbemorefavouredusingRP chromatography.

Itisnotpossibletomakefirmconclusionsonthisissuefrom ourpresentwork,asthec-coefficientsdeducedareamixtureof thecontributionofmasstransferandeddydispersion.Thelatteris assumedtobeindependentofflowinthesimplifiedvanDeemter treatmentusedinthisstudy(Eq.(1))whereasclearlyitisnot.

A further comparison of kinetic performance is available throughkineticplotsasthismethodalsoconsidersthedifferences inviscositybetweentheeluentsusedinHILICandRP(seebelow).

3.4. Effectofflowonretention

Themeasurementofparametersrelatedtosoluteretentionon smallparticlecolumnsiscomplicatedbytheeffectsoffrictional heating andpressure which both increase as flow isincreased. Theseeffectscanbesolutedependent.Indeed,significantchanges inretentionfactorwereobservedforallanalytesinbothRPand HILICmodes(Fig.3),whichshowsthe%lossinretentiongoing fromthelowesttothehighestflowusedfor1.8,3.5andalso5␮m columns,introducedtocompareeffectsatthelowestpressures. Forallcolumns,thelowestflowrateusedwas0.025mL/min.The highestflowratesusedfor1.8␮mparticleevaluationinHILICwas 2.0mL/minand1.0mL/mininRP.Forthe3.5␮mcolumninHILIC, themaximumflowwas2.0mL/min,whereasforRPconditionsthe maximumflowusedwas1.5mL/minand1.7mL/minfor nortripty-lineandnaphthalenedatarespectively.Clearly,theeffectsofhigh flowonretentionaregreatestforthe1.8␮mcolumnandleastfor the5␮mcolumn.Theseeffectsarerarelymentionedinstudiesof columnperformanceasafunctionofflow.

Fig.3showsasmall%lossinkbetweenhighandlowflowfor naphthaleneintheRPmodewhichismostsignificantforthe1.8␮m column.Valuesofkusuallydecreasewithincreasingtemperature

[33]and indeedin columnsofsimilardimensions,temperature increasesofupto20◦Chavebeennotedduetofrictionalheating

Fig.3.Percentagelossinkbetweenthehighestandlowestflowrateused(detailed inSection3.4)for(a)1.8␮m,(b)3.5␮mand(c)5␮mparticles.

500barpressureincrease[35])forthishigherMWionicspecies. Cytosine and5-(hydroxymethyluridine) showlargedecreases in retentionwithincreasedflow.InHILIC,increasedpressurealone appearstoreducek[36,37]andthereforetheeffectsofpressure and temperatureact inthesame direction.For nortriptylinein theHILICmode,increasedtemperatureexceptionallyappearsto increaseretention[10].Thereforepressureandtemperatureeffects balanceoutleadingtothesmallvariationinkwithflowindicated inFig.3.

ItisconceivablethatthevariationsinkwithflowshowninFig.3, especiallyforthe1.8_␮mcolumnmighthaveaffectedthevaluesof thereducedb-andc-coefficientsforthiscolumn.However,results forthesecoefficientsaresimilaronthelargerparticle3.5␮m col-umn,whichshowedmuchsmallervariationsinkwithflow(see

Table4).Thisfindinggivesconfidencethatthecoefficientsdeduced bycurvefittingarenotseriouslyinfluencedbykvariation,orby theeffectsoffrictionalheatingduetothemobilephasepercolation throughthepackedbed.

Fig.4.KineticplotoftoversusNforHILICandreversed-phaseusingapressure

maximum=1000barforthe1.8␮mparticlepackedcolumns.Symbolsareasin

Fig.1.

3.5. KineticplotcomparisonbetweenHILICandRPmodes

TheinfluenceoftheB-termandeluentviscosity()onthe maxi-mumachievablekineticperformanceisestablishedbyconsidering thefollowingexpressionofDesmetetal.[38]:

Nmax=P

v

B

exp

(5)

Operatingalengthofcolumnatafixedpressuremax(P)with a specific bed permeability (Kv) any reduction in both the B-coefficientandviscositywouldaffordhigherachievablemaximum platenumbers.Byincreasingcolumnlengthtoaffordlargerplate numbers,higherpressuresarerequiredtoachievelinearvelocities inexcessoftheoptimumvelocitytolimitthecontributionofthe B-termtoefficiency.ThemoderateB-coefficientsinHILICcombined withthelowviscosityofthemobilephasesthereforeoffers particu-laradvantagesinaccommodatinglongcolumnsforhighresolution analysis.Thelowermobilephaseviscositymightcontribute pos-itivelytofastanalysis,sincehigherlinearflowvelocitiescanbe adoptedwithinthepressurelimitoftheinstrument.Furthermore, forsomesolutesthatcanbeanalysedineitherHILICorRPmode, enhanceddiffusioncoefficientsinHILICshouldfavourablypromote masstransferinfastanalysis.

Fig.5.(a)PlotofLogk_{versusLog[M+]and(b)kversus1/[M+].HILIC}

condi-tionswere94.8%ACNcontaining1–8mMoverallammoniumformatw

wpH3.0and

reversed-phasewas33.4%ACNcontaining1–50mMoverallammoniumformatw w

pH3.0.

However,theachievementofsuchplatecountsentailstheuseof verylongcolumns,thepredictedlengthsbeingaround60cmfor RPconditionsand80cmforHILIC.

Thecurvesalsoindicatetherelativeperformanceatshort anal-ysistimes(lefthandsideoftheplots).Atto=10sthepredictions were19,000,12,600,17,000 and16,800 platesfor naphthalene (RP),nortriptyline(RP),nortriptyline(HILIC)andcytosine(HILIC) respectively.Thepredictedcolumnslengthsforthefastanalysis regionwere around 10cm. The differences in theplate counts relates to those particular column lengths being operated at 1000barandwellintotheC-termdominatedregion.Recallthat naphthaleneinRPhadaverylowC-coefficientwithlittlelossin efficiencyobservedin thehighflow region(Fig.1).Incontrast, thepoorerperformanceofnortriptylineinRPmodeintheC-term regionoftheplotisduetoitslowmobilephasediffusion coeffi-cient.Aswithanykineticplotpredictionscautionmustbeadhered toastheinitialvanDeemterdatawereconstructedusing5cmlong columns.Rutaetal.[39]pointedoutthatdifferencesinthe pack-ingqualitiesoflongercolumns,orfrictionalheatingeffects,make itdifficulttorealiseaccuratelyinpracticethepredictionsmadefor long/coupledcolumns.

3.6. Retentionmechanismelucidation

Inordertodiscovermoreabouttheretentionmechanism in thedifferentseparation modes,particularlywithregardtoionic processes,weinvestigatedtheeffectofbufferstrengthonk.Cox andStout[17]investigatedionicretentionusingplotsofkversus 1/[M+]where[M+]istheconcentrationofthedisplacingbuffer cation ofthe samecharge as thesolute.These plots shouldbe linearfor apurecation-exchangeprocessfor basiccompounds, withthelinepassing through theoriginat infinite buffer con-centration.Ay-axisinterceptoftheplotallowsthecontribution ofnon-ionicprocessestoretentiontobeassessed.Fig.5 shows theplotsforHILIC(94.8%ACNcontaining1–8mMoverall ammo-niumformatew

wpH3.0)andreversed-phase(33.4%ACNcontaining 1–50mMoverallammoniumformatew

wpH3.0).Thecolumn hold-upvolume(Vm)wasdeterminedusingthenon-retainedcompound tolueneinHILIC,anduracilinRP.Thesemarkerswereincorporated ineveryinjectedsample,asVmcandependonmobilephase prop-ertiessuchastheionicstrength[40].Fig.5showsthattheretention ofnortriptylineinHILICdecreasessubstantiallyasthebuffer con-centrationincreases,indicatingthepresenceofionicretention.The curvatureoftheplotissimilartothatfoundinapreviousstudy[7], alsousingabaresilicastationaryphase(Luna,Phenomenex).Itis possiblethiscurvaturecouldbeduetoasmalldegreeofionpairing, whichhasbeensuggestedasfeasibleatleastwiththeacetateanion

[41].Thecurvatureprecludestheaccurateestimationofthe contri-butionofionexchangetoretentionthatwouldbepossiblewitha linearplot,butextrapolationtoinfinitebufferconcentration indi-catesthataveryhighproportionoftheretentionprocessisdueto ionexchange.Incontrast,nortriptylineinRPmodeshowsasmall increaseinretentionwithincreasingbufferstrength,indicatingthe absenceofcationexchangeundertheseconditions.Itispossible thatthepresenceof hydrophobicligandsreducestheacidityof silanolgroups intheRPmodeand thus alsoreducesthe influ-enceofionexchange,asthesestationaryphasesusethesamebase silica.Thesmallincreaseinretentionfornortriptylinewith increas-ingbufferstrengthmightagainbeattributedtoion-paireffects, althoughthesearelikelytobemuchsmallerthaninHILICduetothe lowerconcentrationofacetonitrileintheRPmobilephase. Alter-natively,itispossiblethatsomepositivelychargedgroupsexiston theRPsurfacethatareintroducedinthebondingprocess[42].The repulsiveeffectofsuchgroupswouldbemoderatedwith increas-ingbufferconcentration.Thestrongcomponentofionicretention ofnortriptylineinHILICcouldwellinfluenceitssmallervalueofthe b-coefficientfoundinthismodecomparedwithRP(seeTable4and discussionin3.2),duetoreductionofsurfacediffusioneffects.The weakbasecytosinedoesnotappeartobechargedundertheHILIC conditionsused,asitsretentionincreaseswithincreasingbuffer strength(Fig.5(b)).Increase inretention withincreasingbuffer strengthisalsoshownfortheneutral5-(hydroxymethyl)uridine.It ispossiblethattheincreasedbufferstrengthincreasesthe contri-butionofthepartitionprocesstoretentionforthesecompounds, byincreasingthethicknessofthewaterlayeronthesilicasurface. Clearly,theeffectsofbufferconcentrationhaveonlybeenstudied atthesingleorganicsolventconcentrationusedforeachsolutein thekineticstudies.Itisquitepossiblethatthebalanceofthevarious contributingmechanismschangeswith%organic,andamore com-prehensiveseriesofexperimentswouldbenecessarytoexamine thisquestionfurther.

Insummary,thenatureoftheinteractionsinboth chromato-graphicmodesiscomplex.Itshouldcertainlybeconsideredthat theseinteractionsinfluencethemobilityofanalytesintheadsorbed state.

4. Conclusions

Significantdifferencesin both thereduced and non-reduced axialdiffusionbehaviourofsolutesintheRPandHILICmodeswere observedoncolumnsmadefromthesamesilica.Thesedifferences wereevaluatedusingcurvefittingofthevanDeemterequation todataofplateheightversusflow andbypeakparking experi-ments.Anovelmethodwasemployedusingswitchingflowtoa dummycolumnduringtheparkingstep.Simplenon-reducedplots forthesamebasiccompound(e.g.nortriptyline)ineithermode indicatedthatB-coefficientswerelargerinHILICthanRP,and C-coefficientsweresmaller.Thusthepractitionerwillobservelower efficiencyatlowflowratebuthigherefficiencyathighflowratein HILICcomparedwithRPforsimilarbasiccompounds.Theseresults areasexpected,andattributablemerelytothemuchhighersolute diffusivityofsuchcompoundsinthelowviscosity,highACN con-centrationmobilephasesusedfortheminHILICcomparedwith RP.

phase,increasingtheb-coefficient.Incontrast,inHILIC,localised adsorption(orionicretention)maycontributesignificantlytothe retentionmechanism,atleastforbaresilicacolumnsusingmobile phasesofrelativelylowwatercontent.Surfacediffusioninthelayer ofwateronthesurfaceofHILICcolumnsalsoseemsmuch less likely.Thereducedc-coefficientswerealwayshigherinHILICthan RP.Thiscouldbeduetosloweradsorption–desorptionkineticsin HILIC,butfurtherexperimentsarenecessarytoestablishwhether this,orotherinterpretationsoftheincreasedvalues,aremorelikely

[30].

ThenatureofthemechanisminHILICisill-definedaswasshown bytheeffectofvaryingbufferconcentrationonretention.Ionic retentionofcationsappearstoplayamuchgreaterroleinHILIC thanRP.However,theretentionmechanismmaybesignificantly differentonbondedphaseHILICcolumnscomparedwiththebare silicacolumnsusedinthisstudy.Practitionersshouldbewaryof retentionfactorshiftsathighflowratethatareparticularly appar-entforsomesolutesinHILIC,whereforothersolutes,decreasein retentionbothwithincreasedpressureandwithincreased temper-atureactinthesamedirection.Nevertheless,comparableresults wereobtainedinthepresentstudyforbothsub-2␮mandlarger particlestationaryphases,showingthatthedeductionswerenot grosslyaffected.

Kineticplot analysisshows HILICcangive improved perfor-mance over RPwhen highefficiencies are required,using long columnsat low flow rates. Thisimprovement results fromthe lowviscosityoftypicalHILICmobilephases,allowingtheuseof longercolumns,togetherwiththeeffectoflower B-coefficients thanexpected.Forfastanalysisonshortcolumns,theadvantage oflow mobilephase viscosityin HILICisopposed byhigher C-coefficientsthanmightbeexpected.

Acknowledgment

This work was supported by the United Kingdom Engi-neering and Physical Sciences Research Council [grant number EP/J016578/1].TheauthorsthankAnneMackatAgilent Technolo-giesforexploratoryworkoncomparingHILICandRP.

References

[1]A.J.Alpert,J.Chromatogr.499(1990)177.

[2]A.J.Martin,R.L.Synge,Biochem.J.35(1941)1358.

[3]D.V.McCalley,U.D.Neue,J.Chromatogr.A1192(2008)225.

[4]N.P.Dinh,T.Jonsson,K.Irgum,J.Chromatogr.A1320(2013)33.

[5]S.M.Melnikov,A.Höltzel,A.Seidel-Morgenstern,U.Tallarek,Anal.Chem.83 (2011)2569.

[6]S.M.Melnikov,A.Höltzel,A.Seidel-Morgenstern,U.Tallarek,Angew.Chem. (Int.)51(2012)6251.

[7]D.V.McCalley,J.Chromatogr.A1217(2010)3408.

[8]N.P.Dinh,T.Jonsson,K.Irgum,J.Chromatogr.A1218(2011)5880.

[9]Y.Kawachi,T.Ikegami,H.Takubo,Y.Ikegami,M.Miyamoto,N.Tanaka,J. Chro-matogr.A1218(2011)5903.

[10]A.Kumar,J.C.Heaton,D.V.McCalley,J.Chromatogr.A1276(2013)33.

[11]N.Gray,J.Heaton,A.Musenga,D.A.Cowan,R.S.Plumb,N.W.Smith,J. Chro-matogr.A1289(2013)37.

[12]D.V.McCalley,J.Chromatogr.A1193(2008)85.

[13]J.Ruta,S.Rudaz,D.V.McCalley,J.-L.Veuthey,D.Guillarme,J.Chromatogr.A 1217(2010)8230.

[14]D.V.McCalley,J.Chromatogr.A1171(2007)46.

[15]J.Heaton,N.Gray,D.A.Cowan,R.S.Plumb,C.Legido-Quigley,N.W.Smith,J. Chromatogr.A1228(2012)329.

[16]J.Li,P.W.Carr,Anal.Chem.69(1997)2530.

[17]G.B.Cox,R.W.Stout,J.Chromatogr.384(1987)315.

[18]D.V.McCalley,J.Chromatogr.A1218(2011)2887.

[19]J.H.Knox,H.P.Scott,J.Chromatogr.282(1983)297.

[20]A.Liekens,J.Denayer,G.Desmet,J.Chromatogr.A1218(2011)4406.

[21]D.B.Ludlum,R.C.Warner,H.W.Smith,J.Phys.Chem.66(1962)1540.

[22]P.J.Dunlop,C.N.Pepela,B.J.Steel,J.Am.Chem.Soc.92(1970)6743.

[23]L.R.Snyder,J.J.Kirkland,IntroductiontoModernLiquidChromatography,2nd ed.,Wiley,NewJersey,1979.

[24]J.Li,P.W.Carr,Anal.Chem.69(1997)2193.

[25]F.Gritti,G.Guiochon,J.Chromatogr.A1218(2011)896.

[26]F.Gritti,G.Guiochon,J.Chromatogr.A1297(2013)85.

[27]G.Desmet,K.Broeckhoven,J.DeSmet,S.Deridder,G.V.Baron,P.Gzil,J. Chro-matogr.A1188(2008)171.

[28]K.Miyabe,G.Guiochon,J.Chromatogr.A1217(2010)1713.

[29]F.Gritti,G.Guiochon,J.Chromatogr.A1128(2006)45.

[30]F.Gritti,G.Guiochon,J.Chromatogr.A1302(2013)55.

[31]H.Colin,J.C.Diez-Masa,G.Guiochon,T.Czajkowska,I.Miedziak,J.Chromatogr. 167(1978)41.

[32]R.W.Stout,J.J.DeStefano,L.R.Snyder,J.Chromatogr.282(1983)263.

[33]L.R.Snyder,J.J.Kirkand,J.W.Dolan,IntroductiontoModernLiquid Chromatog-raphy,3rded.,Wiley,NewJersey,2010.

[34]K.Kaczmarski,J.Kostka,W.Zapala,G.Guiochon,J.Chromatogr.A1216(2009) 6575.

[35]M.M.Fallas,U.D.Neue,M.R.Hadley,D.V.McCalley,J.Chromatogr.A1209(2008) 195.

[36]M.M.Fallas,U.D.Neue,M.R.Hadley,D.V.McCalley,J.Chromatogr.A1217(2010) 276.

[37]U.D.Neue,C.J.Hudalla,P.C.Iraneta,J.Sep.Sci.33(2010)838.

[38]G.Desmet,D.Clicq,P.Gzil,Anal.Chem.77(2005)4058.

[39]J.Ruta,D.Zurlino,C.Grivel,S.Heinisch,J.-L.Veuthey,D.Guillarme,J. Chro-matogr.A1228(2012)221.

[40]G.Greco,S.Grosse,T.Letzel,J.Chromatogr.A1235(2012)60.

[41]D.H.Marchand,P.W.Carr,D.V.McCalley,U.D.Neue,J.W.Dolan,L.R.Snyder,J. Chromatogr.A1218(2011)7110.

[42]D.V.McCalley,Anal.Chem.78(2006)2532.