Development of a flow method for the determination of phosphate in estuarine and freshwaters - Comparison of flow cells in spectrophotometric sequential injection analysis

Development

of

a

flow

method

for

the

determination

of

phosphate

in

estuarine

and

freshwaters—Comparison

of

flow

cells

in

spectrophotometric

sequential

injection

analysis

Raquel

B.R.

Mesquita

_,

_M.

_Teresa

_S.O.B.

_Ferreira

_,

_Ildikó

_V.

_Tóth

_,

Adriano

A.

Bordalo

_,

_Ian

_D.

_McKelvie

_,

_António

_O.S.S.

_Rangel

a_{CBQF/EscolaSuperiordeBiotecnologia,UniversidadeCatólicaPortuguesa,R.Dr.AntónioBernardinodeAlmeida,4200-072Porto,Portugal}

b_{LaboratoryofHydrobiology,InstituteofBiomedicalSciencesAbelSalazar(ICBAS)andInstituteofMarineResearch(CIIMAR),}

UniversidadedoPorto,Lg.AbelSalazar2,4099-003Porto,Portugal

c_{REQUIMTE,DepartamentodeQuímica,FaculdadedeFarmácia,UniversidadedePorto,RuaAníbalCunha,164,4050-047Porto,Portugal}

d_{SchoolofChemistry,TheUniversityofMelbourne,Victoria3010,Australia}

Keywords:

Sequentialinjection

Phosphatedetermination

Estuarineandriverwater

Spectrophotometry

Multi-reflectiveflowcell

a

b

s

t

r

a

c

t

Asequentialinjectionsystemwithdualanalyticallinewasdevelopedandappliedinthecomparisonof twodifferentdetectionsystemsviz;aconventionalspectrophotometerwithacommercialflowcell,anda multi-reflectiveflowcellcoupledwithaphotometricdetectorunderthesameexperimentalconditions. Thestudywasbasedonthespectrophotometricdeterminationofphosphateusingthe molybdenum-bluechemistry.Thetwoalternativeflowcellswerecomparedintermsoftheirresponsetovariationof samplesalinity,susceptibilitytointerferencesandtorefractiveindexchanges.Thedevelopedmethodwas appliedtothedeterminationofphosphateinnaturalwaters(estuarine,river,wellandgroundwaters). Theachieveddetectionlimit(0.007mMPO43−)isconsistentwiththerequirementofthetargetwater samples,andawidequantificationrange(0.024–9.5mM)wasachievedusingbothdetectionsystems.

1. Introduction

Phosphateisanimportantroutineparameterinwater analy-sis,beingsimultaneouslyanessentialmacronutrientandapossible pollutant,whenitsconcentrationisabnormallyhigh.The quantifi-cationofphosphateindifferentwaterbodiesisimportantsincethe increaseofphosphateconcentrationsinsurfacewatersisusually linkedtodiffusesources(likeagriculturalrun-offs).

Methods applicable tophosphate determinationin adiverse rangeofnaturalwaters(estuarine,river,wellandgroundwaters) withhighlyvaryingintrinsiccharacteristics,areneeded. Addition-ally,effectivemethodsmusthandletheproblemsarisingfromthe salinity gradient,shouldbefast,withlowreagent consumption andshouldprovidethepossibilityofrealtimeassessment.Flow analysistechniquesmeettheserequirements,whichmakethema valuabletoolformonitoringprocesses.

Flowmethods forthedetermination ofphosphate inwaters have been extensively investigated as listed in comprehensive reviewsbyEstelaandCerdà[1],MotomizuandLi[2]and

Mon-∗ Correspondingauthor.Tel.:+351225580064;fax:+351225090351.

E-mailaddress:aorangel@esb.ucp.pt(A.O.S.S.Rangel).

betandMcKelvie[3].Amongthedifferentflowanalysistechniques proposed inthelast twentyyears, sequential injection analysis (SIA)hasgainedparticularattentionforwateranalysis[4].Samples andreagentsaresequentiallyaspiratedtoaholdingcoil,and subse-quentlytransportedtothedetectorbyflowreversal.Theprotocol sequence,responsiblefor thoseactions,is computercontrolled. SIAmethodshave recognisedmeritsof robustnessand versatil-ityenablingthepossibilitytochangethedeterminationconditions withoutphysicalreconfigurationofthemanifolds.Inthisscenario, severalSIA methods have beenreported for the spectrophoto-metric determination of phosphate [5–20]as listed in Table1. However,thesemethodsweregenerallydevelopedforthe appli-cationtoasingletypeofwater(evenwhenothersampleswere alsoused).Withintheseapplications,someweredesignedtocope withthelowconcentrationoftheanalyteandproposeddistinct strategiestohandlethematrixdifferencesandtheinterferents.For seawatersamples,Maetal.[5]adoptedastrategytodirectly con-centratephosphomolybdeniumbluefromthewatersampletoan HLBcartridge,thiswayminimizingsalinityeffectofthesamples, butcompromisedsamplingfrequency.Forwastewaterandmicro algaemedium[11],amixingchamberwasintroducedintoa man-ifoldtoenhancethemixingconditions,andtoachieveadjustable dilutionofthesamplescoupledwiththestandardaddition

pro-Table 1 Sequential injections systems for the determination of phosphate. Detection system Sample P fraction Reagents Dynamic range RSD LOD Determination rate (h − 1) Reference date Spectrophotometric Seawater Soluble reactive phosphorus Molybdate, ascorbic acid 0.034–1.134 m M 2.50% 1.4 nM 6–10 [5] Fluorimetric Coastal waters Phosphate Molybdate, rhodamine (plus ammonia determination) 0.5–5.0 m M – 0.05 m M 100 [6] Fluorimetric Coastal waters Filterable Reactive phosphate (FRP) Molybdate, rhodamine 0.5–5.0 m M – 0.05 m M 270 [7] Fluorimetric River and marine waters Reactive phosphate Molybdate, rhodamine (plus ammonia determination) 0.3–4.0 m M – 0.3 m M 120 [8] Spectrophotometric Urine Phosphate Molybdate, tin(II) chloride 0.01–30 m M 3.90% – 30 [9] Spectrophotometric Beverages, wastewater, urine Orthophosphate Molybdate U p to 20 mg P L − 1 2.4%, 1.8% 100 m g P L − 1 18 [10] Spectrophotometric Water samples Orthophosphate Molybdate, ascorbic acid used in standard addition 62.5–250 ng P (added) 4% 24 m g P L − 1 14 [11] Turbidimetric Urine Phosphate Calcium carbonate inhibition, calcium phosphate crystallisation 0.1–0.8 m g L − 1; 0.2–1.5 g L − 1 0.97–1.90%; 1.1–2.0% 0.01 mg L − 1; 14 m g L − 1 12; 15 [12] Spectrophotometric Water samples Phosphate Molybdate, ascorbic acid 19.9–196.0 m g L − 1 1.22% 9.92 m g L − 1 54 [13] Spectrophotometric Wastewater Phosphate Molybdate, vanadate (plus Si determination) 0.026–0.485 m M 2.10% 7.4 m M 30 [14] Spectrophotometric Waters and sediments Phosphate Molybdate, ascorbic acid 0.2–7 m g L − 1 – 0.1 mg L − 1 75 [15] Spectrophotometric – Orthophosphate Molybdate, ascorbic acid 4–40 m g L − 1 – 4 m g L − 1 48 [16] Spectrophotometric Natural water and effluent stream Phosphate Molybdate, ascorbic acid 0–70 m g L − 1 0.90% 0.5 mg L − 1 18 [17] Spectrophotometric Tap and well waters Phosphate Molybdate, tin(II) chloride (plus Fe(II) determination) 0.1–1 m g P L − 1 4.10% 20 m g P L − 1 7 [18] Spectrophotometric Wastewater Phosphate Molybdate, vanadate U p to 12 mg P L − 1 <1.4% 0.2 mg P L − 1 23 [19] Spectrophotometric Natural and wastewater Orthophosphate Molybdate, vanadate, malachite green, tin(II) chloride U p to 18.0, 0.4, 4.0 mg P L − 1 2.1, 18, 1.7% 0.15, 0.01, 0.01 m g P L − 1 30 [20]

cedure.Foralakeandatapwatersample,WuandR ˚uˇziˇcka[13] appliedthekineticstoppedflowmeasurementmode,whichcan alsocontributetothereduction ofsamplematrix effectsinthe appliedLab-on-ValveSIAmode.Thesamestoppedflowstrategy wasusedearlierbyMas-Torresetal.[14],tosimultaneously mea-surephosphateandsilicateinwastewaters,althoughundercertain conditionsthemutualinterferenceofthetwoanalytescouldbe detected. Another interesting approach presented by the same researchgroup[18]wasbasedontheuseofthesamplesolutionas carriertodetectmultiplicativeinterferencesinthedetermination. Theuseofthesameflowmanifoldandprotocolsequencefor watersamplesofdifferentoriginsdoesnotusuallyleadtogood qualityresults,duetobothchemicalandphysicalinterferences. Forexample,largedifferencesinsamplesalinitymayinduceshifts in theequilibriumofchemical reactions,andalsomayproduce interferencesinthespectrophotometricdetection,bycreatinglight refractioninthereactioninterfaces,seriouslyaffectingthe analyt-icalsignal[21].

Inthiswork,weproposetheuseofamulti-reflectivecell(MRC) [22]coupledtoasequentialinjectionsystemtotacklethis prob-lem.Intheseconditions,theSIAmanifoldprovidesthenecessary “alacarte”in-linesampletreatmentandtheminiaturized spec-trophotometer(MRCwithaLEDlightsourceanddetector)provides detectionwithminimumschliereneffect.Tohighlightthe mer-itsoftheflowcell,amanifoldthatallowsthecomparisonwitha conventionalflowcell(CFC)placedinaUV–Vis spectrophotome-teris presented.TheSIAmethodologywasbasedonthewidely usedphospho-molybdenumbluereaction.Thisreactionhasbeen reportedashighlysensitiveandselectivewithfewpossible inter-ferences,namelyAs(V),SiandGe[23].Thebestconditionsforthe colorimetricreactionwereassessedandtheminimizationof possi-bleinterferenceswaseffectivelycarriedoutin-line.Thedeveloped system, waseffectively appliedto estuarine (salinitygradient), river,wellandgroundwaters.

2. Experimental

2.1. Reagentsandsolutions

Allsolutionswerepreparedwithanalyticalgradechemicalsand deionisedwater(specificconductancelessthan0.1mScm−1_).

The 20gL−1 _{ascorbicacid solutionwas weekly preparedby} dissolving2.0gofascorbicacid(Normapur,France)in100mLof deionisedwaterandwaskeptinadarkglassflask.

The molybdate reagent was weekly prepared by dissolving 1.6g of ammonium heptamolybdate-tetra-hydrate fromMerck, Germany(16gL−1_{), 40mgofpotassiumantimony(III)oxide} tar-trate hemi-hydrate fromSigma, Germany (0.4gL−1_{) and 0.75g} of tartaric acid from Merck, Germany (7.5gL−1_{) in deionised} water,followedbyadditionof10mLof6Msulphuricacid from Merck,Germany(0.6M).Afterhomogenisation, deionisedwater wasaddedto100mLandthesolutionwaskeptindarkglassflask. Phosphatestocksolution(29.4mM)waspreparedbydissolving 0.40gofpotassiumdi-hydrogenphosphate(fromMerck,Germany) previously dried in 100mL of deionised water and stored in a refrigerator.Phosphateintermediatestandardsolutionsof595mM and11.9mMwerepreparedeveryfortnightbyappropriate dilu-tionofthestocksolutionandstoredintherefrigerator.Working standards,0.024–9.52mM,werepreparedweeklybyappropriate dilutionandstoredintherefrigeratorwhennotinuse.

2.2. Samplecollectionandpreparation

Differentwatersampleswerecollectedandanalysed:estuarine waters(withasalinitygradientalongtheestuary),riverwaters, wellwatersandgroundwater.

Theestuarinewatersampleswerecollectedfromthreedifferent PortugueseestuarieslocatedinNWPortugal:Ave(41.3◦_N,08.7◦_W), Cávado(41.5◦_N,08.7◦_{W)andDouro(41.1}◦_N,08.6◦_W)rivers.For

Fig.1.SIAmanifoldforthespectrophotometricdeterminationofphosphate:molybdatereagent,ammoniumheptamolybdate-tetra-hydrate16gL−1_{,potassiumantimony(III)}

oxidetartratehemi-hydrate0.40gL−1_{,tartaricacid7.5gL}−1_{andsulphuricacid0.60M;ascorbicacid,20gL}−1_{;S,sampleorphosphatestandard;W,waste;PP,peristaltic}

pump;HC,holdingcoil,400cmlong;Ri,reactioncoils,270cmlong;SV,eightportselectionvalve;CFC,spectrophotometerat660nmwithaconventionalflowcell;MRC,

eachestuary,thewaterwascollectedfromthesurfaceatthree dif-ferentlocations:atclosetotheocean(location1)andtwoother upstream (locations 2 and 3), location 3 havingthe less influ-encefromtheocean.(MapsoflocationscanbefoundinElectronic SupportingInformation.)

Sampleswerealsocollectedfromthesamerivers,butfurther upstreamfromtherivermouthtoensurethattheywerenot influ-encedbytheestuary,asindicatedbythemeasuredsalinityvalues. Welland ground waterswere alsocollected inthenorth of Portugal.Allthewatersampleswereintroduceddirectlyintothe SIAsystem.Therefore,resultsonphosphateanalysiscorrespondto thefractionofTotalReactivePhosphorous[24].

2.3. Sequentialinjectionmanifoldandprocedure

The sequential injection manifold used for the colorimetric determinationofphosphatewiththetwopossibledetection sys-temsisdepictedinFig.1.

Solutions were propelled by a Gilson Minipuls 3 peristaltic pump,withPVCpumpingtube.Thiswasconnectedtothe cen-tralchannelofaneightportselectionvalve(ValcoVICI51652-E8). AlltubingconnectingthedifferentcomponentswasmadeofPTFE (Omnifit),with0.8mmi.d.

A personal computer (Samsung SD 700) equipped with a PCL818L interfacecard,runninghomemadesoftware writtenin QuickBasic4.5,wasusedtocontroltheselectionvalvepositionand theperistalticpumpdirectionandspeed.

A conventional spectrophotometer (Hitachi UV-Vis, 100-40) wassetatthewavelengthof660nmandequippedwithaHellma 178.711-QS flow-cell (10mm light path, 30mL inner volume). Whiletheoptimaldetectionwavelengthforascorbicacidreduced phosphomolybdateisca.880nm[25],thespectralabsorptionband isverybroad,anda wavelengthof660nmwasselectedforthe spectrophotometerinordertoallowdirectcomparisonwiththe otherdetectionsystemused,whichwasespeciallydesigned multi-reflectiveflowcell(MRC)previouslydescribedbyEllisetal.[22]. ThelightsourceinthisdetectorwasaredLED(maxat660nm)light sourceconnectedtoa12Vpowersupplyregulatedto5Vusinga multimeter.Theoutputvoltagewassetto0VwhiletheLEDwason andusingdeionisedwater.Fordetaileddescriptionofthedetector thereadersshouldrefertotheoriginalpaper[22].Analyticalsignals wererecordedonaKipp&ZonenBDchartrecorder.

Thesequenceofstepswiththerespectivetimeandvolumesfor thedeterminationofphosphateisshowninTable2.

Thefirststepistheaspirationofsample(stepA),followedby theaspirationofthemolybdatereagentandascorbicacid(stepsB andC).Mixingisthenpromotedbythereversaloftheflowwhile propellingtheplugstowardsthedetector(stepH).

Forthelineardynamicrangeof0.025–0.25mM,fourextrasteps (stepsD–G)wereadded.StepsDandEenabletoobtainadouble plugofsampleand reagent,respectively,andthestepsFandG

resultinastoppingperiodbeforethedetermination.Theflowwas stoppedinthereactioncoilfor30stoensurehighersensitivity.

NoadsorptionofmolybdenumblueonthesurfaceoftheTeflon tubingwasobserved,andnocarryoverbetweenpeakswasdetected undertheworkingconditionspresentedinTable2.Yet,theflow systemwasrinsedwithadiluted(0.05M)NaOHsolutionatthe endofaworkingweek.

2.4. Referenceprocedure

Fortheaccuracyassessment,sampleswereanalysedusingthe vanadomolybdophosphoricacidcolorimetricmethod(APHA 4500-PC)[25].Theassayswereperformedintriplicateandquantified fromstandardcurvesprepareddaily.

3. Resultsanddiscussion

3.1. Studyofthedeterminationparameters

Optimizationofthecolorimetricreactionwascarriedoutusing theUV–Visspectrophotometerwiththeconventionalflowcell.The wavelengthwasselectedto660nm,correspondingtothe maxi-mumemissionoftheLEDtobeusedsubsequently.

3.1.1. Aspirationsequencetotheholdingcoil

Accordingto themolybdenumbluechemistry, theintensely colouredcompoundresults fromthereduction ofthe phospho-molybdicacidbyascorbicacid,APHA4500-PE[25].The phospho-molybdicacidresultsfromthereactionbetweenorthophosphate andthemixtureofammoniummolybdate/potassiumantimonyl tartrateinacidicmedium.

Two possible sequences were investigated: sample–molybdenum solution–ascorbic acid and ascorbic acid–molybdenum solution–sample.The results obtained when the sample was aspirated first showed an 8.5% increase in sensitivity,sothisorderwaschosen.

3.1.2. Reagentconcentration

The molybdate reagent solution is composed of ammo-nium heptamolybdate, potassium antimony(III) oxide tartrate and sulphuric acid. Firstly, the sulphuric acid concentration was fixed and the concentration of heptamolybdate and anti-mony(III)oxidetartratewasstudied.Thistestwascarriedoutby maintainingaconcentrationproportionof40:1(ammonium hep-tamolybdatetetra-hydrate:potassiumantimony(III)oxidetartrate hemi-hydrate)asreportedbyMoraisetal.[11].Theconcentration ofammoniumheptamolybdatetetra-hydratewasvariedfrom8to 18gL−1_{,resultinginaconcentrationrangeof0.20–0.45gL}−1 _for theantimony(III)compound.Thehighestsensitivitywasobserved for 16gL−1 _{of ammonium heptamolybdate tetra-hydrate with}

Table2

Protocolsequenceforthedeterminationofphosphate.

Step SVposition Time(s) Flowrate(mLmin−1_{) Pumpdirection Volume(mL) Action}

A 1 12/7.5a _{3.9 a 780/488}a _{Aspirationofsample/standard}

B 2 1.5 2.9 a 72 Aspirationofmolybdatereagent

C 3 3.5 3.9 a 228 Aspirationofascorbicacid

Db _{4 1.5 2.9 a 72 Aspirationofmolybdatereagent}

Eb _{5 7.5 3.9 a 488 Aspirationofsample/standard}

Fb _6/7c _{17 3.9 b 1105 Propeltoreactioncoil}

Gb _6/7c _{30 0 – 0 Stopflow}

H 6/7c _{70 3.9 b 4550 Propeltodetector(l=660nm)andsystemwashing}

a_{Timeandrespectivevolumeforthelineardynamicrange0.025–0.25mM.}

b_{Stepsaddedincaseofthedynamicrange0.025–0.25mM.}

Table3

Summaryofthechemicalandphysicalparametersstudies.

Parameter Studiedrange Selectedcondition

[(NH4)6Mo7O24.4H2O)](gL−1) 8–18 16 [C4H4KO7Sb](gL−1) 0.20–0.45 0.40 [H2SO4](M) 0.45–0.90 0.45 [C6H8O6](gL−1) 10–50 20 VMolybdatereagent(mL) 51–163 72 Vascorbicacid(mL) 195–325 228 Vsample(mL) 488–910 780

0.40gL−1_{oftheantimony(III)compoundsothesereagent} condi-tionswerechosen.

Theconcentrationofsulphuricacidwasalsostudied;theacidis requiredtoensurethatthepHwas<1inthefinalreactionmixture aswellastoincreasethesolubilityofthereagentcomponents. Con-centrationsof0.45,0.60and0.90Mofsulphuricacidweretested. Withaconcentrationof0.45Mitwasnotpossibletodissolvethe reagents,and aminimumconcentrationof0.60Mwasrequired toobtainfulldissolution.Higheracidconcentrationsresultedina decreaseinsensitivity,so0.60Mofsulphuricacidwasthechosen concentration.

The ascorbicacid concentrationwasalsostudiedwithinthe range 10–50gL−1 _{and the highest sensitivity wasobtained for} 20gL−1_{,sothatwastheconcentrationchosen.Asummaryofall} studiedparametersandthechosenvaluesisshowninTable3. 3.1.3. Lineardynamicrange0.25–9M

Forobtainingthemaximumsensitivityinthisapplicationrange, thestudywasinitiatedbytestingtheinfluenceofthevolumeofthe molybdatereagent(samplevolumesetto488mL,withintherange 1–4mM).Thetestedvolumeswere:51,72,98,and163mL.Avolume of72mLgaveincreasedsensitivityandbetterlinearity,andwasthe chosenvolume.

Optimizationoftheascorbicacidvolumewasperformedwithin therange195–325mL.Thevolumeofascorbicacidwhichprovided thehighestsensitivitywas228mL,sothiswasthevolumechosen. Thisresultcanbeexplainedbythelimitedoverlapping(ormutual dispersion)betweenthereagentandsamplesolutionsinducedby theincreaseoftheascorbicacidvolumeintroducedbetweenthem. Thesamplevolumewasstudiedandavolumeof780mLwas chosenfromtherange488–910mLasitgavethehighestsensitivity. Withthisreassessmentofsamplevolume,alineardynamicrange of0.25–9.5mMwasobtained.

3.1.4. Lineardynamicrangeupto0.25M

Considering that someuncontaminated natural waters may have phosphate concentrations below 0.25mM [26], different approachesweretestedtoachievealowerlimitofquantification. 3.1.4.1. Effectofdifferentreagentsequences—thedoublesampleplug. Toobtainalineardynamicrangeof0.05–0.25mMitisnecessaryto maximizethevolumeofsampleand,atthesametime,to guaran-teeefficientmixingwiththereagents.Toachievebothconditions, thesamplewasintroducedastwoplugs,withoneoneitherside ofthereagent plugs.Aprogram sequenceincludingthedouble plugsforbothsampleandmolybdatereagentwastested.Theorder ofaspirationwiththetwo extraplugswas: sample–molybdate reagent–ascorbic acid–molybdate reagent–sample. The sample volumeofeachplugwassettoabout500mLandthevolumesof molybdatereagentandascorbicacidweremaintainedatthesame valuesderivedfromthepreviousstudy(Section3.1.3).

3.1.4.2. Effect of reaction time—stop-flow. In order to further increasesensitivity,anincreaseinthereactiontimewastested. Aftertheaspirationof alltheplugs,themixturewaspropelled towardsthedetector and thenstopped in thereactioncoil. To ensurethatalltheplugswereinthereactioncoil,thecoillength wassetto270cm,avaluethatwasobtainedfromprevioustests employingadyesolution.Severalstoptimesweretestedranging from0to90sand astoptimeof30swaschosen.While sensi-tivityincreasedwiththeincreasedstoptime,thelinearityofthe calibrationcurvedecreased,soa30sstoptimewasselectedasa compromisebetweensensitivityandlinearityofresponse.

Undertheoptimizedconditionswiththedoublesampleplug setupitwaspossibletoreachthedetectionlimitof0.007mM,and thequantificationlimitof0.023mM.

3.2. Comparisonofthetwopossibledetectionsystems

Theconditionsobtainedfrompreviousstudies(Section3.1)for thelinear dynamicrange of 0.25–9.5mM phosphatewereused (unlessotherwisestated)inordertocomparetherelativemerits ofthetwodetectioncells.

3.2.1. Schliereneffect

Inflowanalysis,thephenomenonknownastheschliereneffect, results fromthedeflectionofthelight beam(signal refraction) causedbythecreatedconcentrationgradient[21]iftherefractive indices of carrier sample and reagents are significantly

differ-Fig.2. SIAtracesshowingacomparisonbetweenthesample(s)andthedeionisedwater(w)peaks:(a)withtheMRC,(b)withtheCFC;andcomparisonofapartialcalibration

ent. Theinterface between thesesolutions canproduce optical lenses andsoa signalresultingoflight deflectionisregistered, schliereneffect.Ifthissignalisconcomitantwiththeonetobe mea-suredatthesamewavelength(lightabsorption),erraticresultsare obtained,withmoreevidenteffectsatlowanalyteconcentrations. This problem is particularly important in SIAsystems because, unlikeFIAmanifolds,noconfluencepointsareusedtoimprovethe mixing.Differentstrategieswereeffectivelyappliedtominimize theimpactofthisphenomenonontheaccuracyandprecisionof spectrophotometricdeterminations,Diasetal.suggestedtheuse of dualwavelength measurement[21] and McKelvieet al.[27] appliedthematrix matchingstrategyforsalinitycompensation. TheMRCwasspeciallydesignedbyEllisetal.[22]tominimize theschliereneffect.ToassessthiscapacityunderSIAbasedflow conditions,sampleandblankwereinjectedandthesignals (ampli-fiedviewinFig.2aandb)werecompared.Theresultspresented onthefigurecorrespondtothedoubleplugprogram(lowerlinear dynamicrange)asthisparticulareffectismorepronouncedwhen increasednumbersofplugsareinjected.SIApeakswereproduced usinganMRCandtheconventionalflowcell,anditcanbeobserved that,whenusingaCFC,thesignalsaremuchmorepronetoerratic signalsclosetothebaseline(Fig.2candd).

TheseresultsindicatethattheSIA-MRCcanbeusedinmatrices withquitedifferentrefraction indices,withnosignificant opti-calinterferenceinthemainsignal.Additionally,theuseofaMRC contributestotheminiaturizationoftheoverallapparatus.This detection system proved to be advantageous for theestuarine watersamples.

3.2.2. Chemicalinterferences

Distinctivelyfromotherspectrophotometricor chemilumino-metricdeterminationsofphosphate,themolybdenumbluemethod isonlysubjecttoasmallvarietyofinterferencesunderflowanalysis conditions.Interferingspeciesinthedeterminationofphosphorous bythephosphomolybdeniumbluemethodareAs(V),SiandGe[23], whichalsoreactwithmolybdatetoformthecorrespondingacids whicharereducedtotherespectiveheteropolybluesbyascorbic acid.

Consideringthattheproposedmethodwasdirectedatnatural waters(estuarine,riverandwellwaters)themostlikely interfer-encewouldbefromsilica.Thisisachemicalinterferencetherefore itwasnotexpectedtoproducedifferentbehaviourinthetwo detec-tionsystem.

Thisinterferencewasassessedatphosphateconcentrationsof 0.25mM,withdifferentamountsofsilicateadded.UsingtheCFC detectionsystem,therewasnosignificantinterferenceuptoa sil-icateconcentrationof32.9mM,howeverwiththeMRCdetection systemsignificant(>5%)interferencewasfoundinthepresenceof 1.3mMsilicate.Theinterference,causedbytheformationof

molyb-dosilicateheteropolyacidcomplexatreducedpHcanexplainthis effect,which becomesmorepronouncedusingthedetectorcell withhighersensitivity.Estuarinewatersareexpectedtobeaffected byseawaterwithvaluesupto60mMofsilica[28].Thepresenceof tartaricacidhasbeenreportedtoeffectivelyreducesilicate inter-ference[23]soitwasaddedtothemolybdatereagent.Different concentrationsoftartaricacidwereused(1.7,3.2,6.4,7.5gL−1_)for asilicateconcentrationof65.7mM.Asexpected,withanincrease inthetartaricacidconcentration,therewasadecreaseofthe per-centageofinterference.Thepercentageinterferencedroppedfrom 18%,withnotartaricacid,to2%,withthe7.5gL−1_{tartaricacid.}

So,eventhoughonlysomeestuarinesampleswerelikelytohave silicatevaluesupto60mMofsilica,7.5gL−1_{oftartaricacidwas} includedinthemolybdatereagent.Thereasonforaddingthe tar-taricacidtothemolybdatereagentinsteadoftotheascorbicacid solutionwasbasedontheaspirationorder,toensurethatitcontacts directlywiththesample.

Thepossibleinterferencefromnitratewasalsostudied,atthe concentrationlevelof200mMofnitrate,andresultedin<3% inter-ferenceonthesignalofa 0.25mMphosphatestandard solution. Thisnitrateconcentrationisaboutthemaximumfoundinthese estuarinewaters[29].

3.2.3. Salinityinterference

Astheaimofthedevelopedmethodwastheapplicationtowater sampleswithawidesalinityrange,i.e.:estuarinewaters,astudyof salinityinterferencewasofhighpriority.Thesalinityinterference wasassessedbycomparingcalibrationcurvesusingstandard solu-tionswithdifferentsalinityvalues;thesesolutionswereprepared byaddingsodiumchloridetothestandards.Thesalinityvalues wereadjustedtoanintermediateandamaximumlevelof salin-ityinestuarinewaters,of9‰and19‰,respectively[30],although itisrecognisedthatsalinitiesat35‰canbefoundinestuariesatthe seawardextent,beingalsodependentonthetidalriverflow con-ditions.Thecalibrationcurvewithpurestandardswascompared tothecalibration curvewithstandardswithadjustedsalinities. Theestimatedslopesofthecurveswereassessedattheconfidence intervalsat95%.Thequalityoftheregressionwastestedby resid-ualanalysis(i.e.randomnessandnormality)andbythecoefficient ofdetermination(i.e.R2_{,whichwasabove0.981inallcases).For} theCFCthereisonlyoverlappingofthelimitswhilefortheMRC thereisanalmostperfect overlappingoftheentireinterval. So thesalinityaffectsmorethespectrophotometricdetectionwiththe conventionalcell(Fig.S1inElectronicSupportingInformation). 3.3. FeaturesofthedevelopedSIAmethod

Asummaryofthemainanalyticalfeatures:thedynamic concen-trationrangesforthetypicalcalibrationcurves,limitsofdetection

Table4

FeaturesofthedevelopedSIAmethodwithbothdetectionsystems.

Detectionsystem Lineardynamic

range(mM)

Calibrationcurvea _{LOD(mM) LOQ(mM) RSD(%)}b _{Determination}

rate(h−1₎ Effluent volume (mL) Spectrophotometer andconventional flowcell 0.30–9.50 A=0.0869 (±0.0035) [PO43−]−0.039 (±0.025); R2_{=0.999(±0.001)} 0.089 0.297 0.6% (2.47±0.02) 0.6% (7.33±0.04) 37 4.6 Multi-reflective flowcell 0.25–7.10 Hc_=15.2(±0.5) [PO43−]+2.6 (±1.4);R2_=0.998 (±0.001) 0.066 0.219 1.3% (2.92±0.04) 0.2% (7.04±0.01) 37 4.6

a_{Valuesinbracketscorrespondtothestandarddeviation,n=5,oftheequationparameters(interdayprecision).}

b_{Valuesinbracketscorrespondtotheconcentrationandthestandarddeviation,n=10(intradayprecision).}

(LOD) and quantification (LOQ),repeatability (RSD), determina-tionrateandeffluentproduction,forthetwodetectionsystems, isshowninTable4.

Thetypicalcalibrationcurvescorrespondtoameanoffour cal-ibration curves withthestandard errors betweenbrackets.The LODandLOQofbothlineardynamicrangeswerecalculatedasthe concentrationcorrespondingtothreeandtentimesthestandard deviationoftheblank,respectively,accordingtoIUPAC recommen-dation[31,32].

The repeatabilitywasassessedby calculationofthe relative standarddeviation(RSD)obtainedbythemeanoftenconsecutive injectionsofanestuarinewatersample[31].

Thedeterminationratewascalculatedbasedonthetimespent percycle.Acompleteanalyticalcyclemusttakeintoaccountnot onlythesumoftimesintheprotocolsequencebutalsothetime requiredtochangethevalvepositionandtoactivatethepump.The analyticalcycletook2.6minforthedoubleplugmethod(lower lin-eardynamicrange)and1.6minforthesingleplugmethod(higher lineardynamicrange).

The mostsignificantconclusion that canbedrawnfromthe observations in Table 4is that lower detection and quantifica-tionlimitsareobtainedwiththemulti-reflectivecell.Thismaybe attributedtothehigherblankvalueandanincreaseintheblank variabilityobtainedwhenusingthespectrophotometerequipped withthecommercialflowcell(Fig.2).Thedetectionlimitcanbe furtherdecreasedifneededwiththemulti-reflectivecellandusing thedoublesampleplugstrategypresentedSection3.1.4.

Thereagentconsumptionvaluesforthesingleplugmethod val-ueswere:1.52mgofammoniumheptamolybdate-tetra-hydrate; 28.8mgofpotassiumantimony(III)oxidetartratehemi-hydrate; 0.54mgtartaricacid;4.24mgsulphuricacidand4.56mgascorbic acid.

3.4. Applicationtowatersamples

ThedevelopedSImethodwasusedfortheanalysisofavariety ofnaturalwaters,withdifferentsourcesandcharacteristics,and theresultswerecomparedtotheonesobtainedbythereference procedure.

3.4.1. Phosphatedeterminationinnaturalwaters

Theestuarineandriverwatersampleswerecollectedin differ-entlocationsoftheAveriver,asmentionedinSection2.2.Someof thosesampleswereconsideredestuarinewatersduetotheir prox-imitytotherivermouth;onewasconsideredriverwaterduetothe

distancetothesea.Thereforethesalinityvaluesweresignificantly differentrangingfrom0.4‰to7.2‰.

Somewelland ground waterswere alsoanalysed and their salinityvalueswerenegligibleandwerenotlisted.Thevaluesare reportedin Table5 asmeanwiththestandard deviation(three replicatedeterminationsforeachsample).

TheFtestindicatesthatthevariancedetectedinthereference methodinmostcasesisgreaterthantheonefromtheproposed methods.Therefore, unequal variances wereconsidered[33] in thet-testforthecomparisonofthemeans. Calculated|t|values inTableS1(ESI)forthetwodetectionsystemsindicatethatthe resultsobtainedusingtheMRCbasedSIAsystemshowedabetter overallagreementwiththereferencemethod.

Toevaluateaccuracy,alinearrelationshipbetweentheresults obtained with the reference procedure and each one of the two possibledetection systems,CFC and MRC,was established (Fig.S2intheElectronicSupportingInformation).Theregression of CCFC (mM) vs CRef.Met. (mM) resulted in the linearequation: CCFC=0.930(±0.196)×CRef.Met.−0.004(±1.199),R=0.953.The lin-earrelationshipbetweenCMRC(mM)andCRef.Met.(mM)resulted intheequation:CMRC=0.966(±0.133)×CRef.Met.+0.185(±0.811), R=0.979.Thevaluesinparenthesisrepresentthe95%confidence limits.Inbothcasesthefiguresshowthattheestimatedslopeand interceptdonotdifferstatisticallyfromvalues1and0,respectively. Therefore,noevidenceofsystematicdifferencescanbepointedout betweenthetwosetsofresults[33].

3.4.2. Recoverystudiesforseveraldifferenttypesofwater samples

Tofurtherassesstheefficacyofthedevelopedsystem, recov-erystudieswereperformedonwatersamplespreviouslycollected fromotherestuariesinthenorthofPortugal(Section2.2)andstored frozen.

Sampleswerespikedwithvolumesof100mL,120mL,150mL and160mLofphosphatestandardsolution(595mM)wereadded to10or20mLofsample.Thecalculationoftherecovery percent-agewasmadeaccordingtoIUPAC[34].Theanalysiswascarried outusingbothdetectionsystems(TableS1inElectronic Support-ingInformation).

TheSIAmethodologyprovidedrecoveryratioswithanaverage of98%(standarddeviation6.1%)and100%(standarddeviation5.5%) fortheCFCandMRC,respectively.Astatisticaltest(t-test)showed thatfora95%significanceleveltherecoveryvaluesdidnotdiffer from100%asthecalculatedt-valueswere1.639and0.096with correspondentcriticalvalues2.475and2.445fortheCFCandMRC,

Table5

Applicationofthedevelopedsequentialinjectionmethodwiththetwopossibledetections,theconventionalflowcellwiththespectrophotometer(SIA-CFC)andthe

multi-reflectivecell(SIA-MRC)tothephosphatedeterminationindifferentwatersamplesandcomparisonwiththereferenceprocedure(Ref.Met.);SD,standarddeviationfrom

3replicas,RDrelativedeviation.

Sampletype SampleID Salinity(‰) Ref.Met. SIA-CFC SIA-MRC

[PO43−]±SD(mM) [PO43−]±SD(mM) RD(%) [PO43−]±SD(mM) RD(%)

Estuarinewater(Averiver) A1 5.9 6.94±0.14 5.93±0.05 −14.6 6.72±0.02 −3.2

A2 1.2 6.13±0.28 5.87±0.03 −4.2 6.16±0.01 0.5 A3 0.1 6.53±0.01 6.34±0.05 −2.9 6.35±0.01 −2.8 A4 7.2 4.75±0.14 3.82±0.02 −19.5 4.71±0.03 −0.7 A5 0.8 5.89±0.14 5.60±0.03 −4.9 5.86±0.02 −0.5 A6 0.1 6.54±0.24 6.31±0.05 −3.5 6.32±0.01 −3.3 A7 6.5 7.84±0.14 7.40±0.05 −5.6 8.05±0.01 2.7 A8 1.1 5.40±0.14 5.59±0.01 3.5 5.75±0.02 6.6 A9 0.5 6.94±0.14 6.82±0.04 −1.8 7.40±0.01 6.6 Groundwater G1 – 3.53±0.01 3.36±0.06 −4.8 3.54±0.03 0.3 G2 – 8.41±0.14 7.30±0.03 −13.2 7.88±0.01 −6.2 Wellwater W0 – 4.23±0.15 4.35±0.04 2.9 4.58±0.01 8.3 Riverwater R0 0.4 4.10±0.01 3.05±0.02 −25.6 3.68±0.01 −10.3

respectively,thusindicatingtheabsenceofmultiplicativematrix interference[33].

4. Conclusions

Thedevelopedmethodenabledtheevaluationoftwo alterna-tivedetectioncellsunderthesameexperimentalconditions,and resultedinaSIAprocedureforphosphatedeterminationapplicable todifferenttypesofwaterwithhighsalinityandlargematrix vari-ability.Theuseofasequentialinjectiontechnique,tofullyexplore itsversatility,provedtobeasuitablechoice.Theapproachofusing twoprotocolsequences,resultingintwolineardynamicranges, canbeappliedtosampleswithtraceconcentrationoftheanalyte. Incomparing thetwodetectionsystems,themulti-reflective flowcellprovedtobebetterforlowerphosphateconcentrations withanachievabledetectionlimitof0.007mM.

Withrespecttothepossibleinterferencesofsilicate,the inter-ference was effectively minimized, in-line, by adding tartaric acid to themolybdenum reagent. The percentage interference, observed for 68mM of silicate, decreased from18% to 2% asa result.

Whencomparedtopreviouslydescribedproceduresbasedon thesameprinciples(Table1),themostsignificantadvantageofthe workdescribedistheuseofasinglemanifoldforthedetermination ofphosphateindifferentmatrices(well,estuarine,river,andsea waters).Theachievabledetectionlimitcomparesfavourablytothe previouslypublishedSIalternatives(Table1)withouttheneedfor apre-concentrationprocedure.Theanalysedwatersampleswere comparedtoareferenceprocedureandprovedtheaccuracyofthe developedmethodforbothdetectionsystems.

Acknowledgements

R.B.R.MesquitathankstoFSEandFundac¸ãoparaaCiênciae aTecnologia(FCT,Portugal)thegrantSFRH/BPD/41859/2007.I.V. Tóth thanksFSEand MinistériodaCiência, TecnologiaeEnsino Superior(MCTES)forthefinancialsupportthroughthePOPH-QREN program.TheauthorsthankFCTfinancialsupportthroughproject PTDC/AMB/64441/2006,andPeterEllis,MonashUniversity,forthe kindprovisionofthemulti-reflectionflowcell.

AppendixA. Supplementarydata

Supplementarydataassociatedwiththisarticlecanbefound,in theonlineversion,atdoi:10.1016/j.aca.2011.06.002.

References

[1] J.M.Estela,V.Cerdà,Talanta66(2005)307–331.

[2] S.Motomizu,Z.-H.Li,Talanta66(2005)332–340.

[3] P.Monbet,I.D.McKelvie,Phosphates,in:L.M.L.Nollet(Ed.),HandbookofWater

Analysis,CRCPress,NewYork,Basel,2007,pp.219–252.

[4] R.B.R.Mesquita,A.O.S.S.Rangel,Anal.Chim.Acta648(2009)7–22.

[5] J.Ma,D.Yuan,Y.Liang,Mar.Chem.111(2008)151–159.

[6]C. Franck, F. Schroeder, J. Automat. Meth. Manag. Chem. (2007),

doi:10.1155/2007/49535,ID49535.

[7] C.Franck,F.Schroeder,R.Ebinghaus,W.Ruck,Talanta70(2006)513–517.

[8] C.Franck,F.Schroeder,R.Ebinghaus,W.Ruck,Microchim.Acta154(2006)

31–34.

[9] D.G.Themelis,A.Economou,A.Tsiomlektsis,P.D.Tzanavaras,Anal.Biochem.

330(2004)193–198.

[10] F.Mas-Torres,J.M.Estela,M.Miró,A.Cladera,V.Cerdà,Anal.Chim.Acta510

(2004)61–68.

[11]I.P.A.Morais,M.R.S.Souto,A.O.S.S.Rangel,J.FlowInjectionAnal.20(2003)

187–192.

[12]B.M.Simonet,F.Grases,J.G.March,FreseniusJ.Anal.Chem.369(2001)96–102.

[13]C.-H.Wu,J.R ˚uˇziˇcka,Analyst126(2001)1947–1952.

[14] F.Mas-Torres,A.Munõz,J.M.Estela,V.Cerdà,Int.J.Environ.Anal.Chem.77

(2000)185–202.

[15]C.X.Galhardo,J.C.Masini,Anal.Chim.Acta417(2000)191–200.

[16]J.R ˚uˇziˇcka,Analyst125(2000)1053–1060.

[17] J.F.vanStaden,R.E.Taljaard,Mikrochim.Acta128(1998)223–228.

[18]F.Mas-Torres,A.Cladera,J.M.Estela,V.Cerdà,Analyst123(1998)1541–1546.

[19]F.Mas-Torres,A.Munõz,J.M.Estela,V.Cerdà,Analyst122(1997)1033–1038.

[20]A.Munõz,F.Mas-Torres,J.M.Estela,V.Cerdà,Anal.Chim.Acta350(1997)

21–29.

[21]A.C.B. Dias, E.P. Borges, E.A.G. Zagatto, P.J. Worsfold,Talanta 68 (2006)

1076–1082.

[22]P.S.Ellis,A.J.Lyddy-Meaney,P.J.Worsfold,I.D.McKelvie,Anal.Chim.Acta499

(2003)81–89.

[23]Z.Marczenko,M.Balcerzak,SeparationPreconcentrationand

Spectrophotom-etryinInorganicAnalysis.Chapter37.Phosphorus,Elsevier,Amsterdam,2000,

pp.326–333.

[24] P.J.Worsfold,L.J.Gimbert,U.Mankasingh,O.N.Omaka,G.Hanrahan,P.C.F.C.

Gardolinski,P.M.Haygarth,B.L.Turner,M.J.Keith-Roach,I.D.McKelvie,Talanta

66(2005)273–293.

[25]APHA-AWWA-WPCF,StandardMethodsfortheExaminationofWaterand

Wastewater,20thed.,AmericanPublicHealthAssociation,Washington,DC,

1998(4500-P).

[26]M.Yaqoob,A.Nabi,P.J.Worsfold,Anal.Chim.Acta510(2004)213–218.

[27]I.D.McKelvie,D.M.W.Peat,G.P.Matthews,P.J.Worsfold,Anal.Chim.Acta351

(1997)265–271.

[28] M.Lacombe,V.Garc¸on,M.Comtat,L.Oriol,J.Sudre,D.Thouron,N.LeBris,C.

Provost,Mar.Chem.106(2007)489–497.

[29]R.B.R.Mesquita,M.T.S.O.B.Ferreira,R.L.A.Segundo,C.F.C.P.Teixeira,A.A.

Bor-dalo,A.O.S.S.Rangel,Anal.Methods1(2009)195–202.

[30] E.L.Lewis,IEEEJ.OceanicEng.OE-5(1980)3–8.

[31]InternationalUnionofPureandAppliedChemistry,Anal.Chem.48(1976)

2294–2296.

[32] InternationalUnionofPureandAppliedChemistry,PureAppl.Chem.67(1995)

1699–1723.

[33]J.C.Miller,J.N.Miller,StatisticsforAnalyticalChemistry,3rded.,EllisHorwood,

NewYork,1993.

[34]InternationalUnionofPureandAppliedChemistry,PureAppl.Chem.74(2002)