Analytica Chimica Acta 571 (2006) 184–190

Flow injection analysis of ultratrace orthophosphate in seawater withsolid-phase enrichment and luminol chemiluminescence detection

Ying Liang, Dongxing Yuan ∗, Quanlong Li, Qingmei LinState Key Laboratory of Marine Environmental Science, Xiamen University, Xiamen 361005, PR China

Received 28 January 2006; received in revised form 13 April 2006; accepted 18 April 2006Available online 27 April 2006

Abstract

Solid-phase extraction technique had been applied to extract molybdophosphoric heteropoly acid (MoP) paired with cetyltrimethylammoniumbromide (CTAB) from seawater matrix using C18 sorbent. Chemiluminescence emission could be generated via MoP reaction with alkalineluminol. Based on these, a novel on-line solid-phase extraction method coupled with flow injection (FI) analysis and luminol chemiluminescencedetection had been established to determine ultratrace orthophosphate in seawater. The MoP–CTAB compound could be efficiently extracted onan in-line Sep-Pak C18 cartridge, and rapidly eluted by 0.3 mol l−1 sulphuric acid–ethanol solution. Then the compound was reduced by luminoltu0rac(p©

Kl

1

ficioti

oi(

0d

o produce chemiluminescence light, which could be detected using a luminescence analyzer. Experimental parameters were optimized using anivariate experimental design. Using artificial seawater with salinity of 35 as a matrix, the standard curve with a linear range between 0.005 and.194 �mol l−1 had been obtained, and the recovery and the detection limit of the proposed method were found to be 92.5% and 0.002 �mol l−1,espectively. The relative standard deviation (R.S.D.), which was determined over eight hour, was 4.66% (n = 7) for the artificial seawater atconcentration of 0.097 �mol l−1 orthophosphate. Si of 200 �mol l−1 would not interfere with the detection of 0.012 �mol l−1 orthophosphate

ompound. Three typical seawater samples were analyzed using both the proposed method and the magnesium hydroxide-induced coprecipitationMAGIC) method, and the results of the two methods showed no significant difference using the t test. Compared to the MAGIC method, theroposed method was more sensitive, time saving and easy for on-line analysis.

2006 Published by Elsevier B.V.

eywords: Sep-Pak C18; On-line solid-phase extraction; Orthophosphate; Molybdophosphoric heteropoly acid; Cetyltrimethylammonium bromide; Luminol chemi-uminescence

. Introduction

Phosphorus is well known as an essential nutrient elementor organisms in aquatic environments [1–3]. Orthophosphates the major form of phosphorus in the phosphorus cycle, espe-ially in natural waters. The biogeochemical cycle of phosphoruss one of the main research topics in chemical and biologicalceanography, and accurately measurement of the concentra-ion of orthophosphate in various water samples is thus verymportant.

The common analysis of orthophosphate in seawater is basedn the quantitative reaction of orthophosphate and molybdaten acidic medium to form molybdophosphoric heteropoly acidMoP), followed by the reduction with ascorbic acid to yield

∗ Corresponding author. Tel.: +86 592 218 4820; fax: +86 592 218 0655.E-mail address: yuandx@xmu.edu.cn (D. Yuan).

phosphomolybdenum blue (PMB) in the presence of antimony[4,5]. Because of its simplicity, rapidity, low interferences, andhigh precision, this determination method for orthophosphatehas been standardized and widely accepted. The detection limitof the method can be as low as sub-micromolar levels. However,to meet the requirements of the determination of oligotrophicopen-ocean waters such as those in the surface of the South ChinaSea, Mediterranean Sea, the subtropical North Pacific Oceanand the Western North Atlantic Ocean where orthophosphateconcentrations are down to nanomolar levels and the standardmethod has failed [6–9], more sensitive and precious methodsare urgently needed.

Much effort has been devoted and many methods have beenreported to detect low-levels of orthophosphate. Those methodprincipals are mostly to concentrate the analyte and extenddetection cell pathlength. The reported techniques included theuse of organic solvents to extract and concentrate from aque-ous solutions the molybdophosphate with malachite green as a

003-2670/$ – see front matter © 2006 Published by Elsevier B.V.oi:10.1016/j.aca.2006.04.036

Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190 185

counter-ion [10]; the magnesium hydroxide-induced coprecipi-tation (MAGIC) method [11]; the luminol chemiluminescencemethod [12]; long pathlength capillary cell absorption spectrom-etry [13]; and the HPLC method [14]. Among these, the MAGICmethod has been widely accepted by oceangraphy researchersdue to its sensitivity and precision. However, with the detectionlimit down to nanomolar levels, this method requires large sam-ple volumes, is highly labor-intense, takes a long time, and isnot amenable to automatic analysis.

Chemiluminescence (CL) is the production of light by achemical reaction. The advantages of CL technique includehigh sensitivity, a wide linear dynamic range and simple instru-mentation. Determinations of ultratrace orthophosphate in freshwater have been achieved by application of the CL techniquebased on oxidation of 3-aminophthalhydrazide (luminol) withvanadomolybdophosphoric heteropoly acid (VMoP) [12] orMoP [15,16]. Zui and Birks [12] reported a solid-phase CLtechnique where VMoP of 150 ml sample was adsorbed andpreconcentrated with the presence or absence of cationic sur-factant. The detection limits were 0.02 �g P l−1 (0.6 nmol l−1)and 0.1 �g P l−1 (3.2 nmol l−1) for those with and without sur-factant, respectively. Another CL method based on the formationof VMoP [15] adopted a flow-through solid-phase based opticalsensor with the detection limit of 4 �g P l−1 (129.4 nmol l−1)for a 1.8 ml sample. A simple and rapid flow injection methodwas developed on the oxidation of luminol by MoP, and thedpptttwpui

h[maohftccottd

t(tiC

The CL reactions are shown as follows [16]:

PO43− + 12MoO4

2− + 27H+ → H3PO4(MoO3)12

H3PO4(MoO3)12 + luminolOH−−→light

2. Experimental

2.1. Reagents and solutions

All the chemicals used in this study were of analyti-cal grade, supplied by Sinopharm Chemical Reagent Co.,China, unless stated otherwise. All solutions were preparedin Milli-Q water. Ammonium heptamolybdate tetrahydrate((NH4)6Mo7O24·4H2O) was recrystallized from ethanol toremove possible remaining orthophosphate and other impuri-ties. A stock solution of color developing reagent was pre-pared by mixing 200 ml of 65 g l−1 ammonium heptamolybdatetetrahydrate solution prepared from the recrystallized chemi-cal and 300 ml of 9 mol l−1 H2SO4 (GR). The solution wasstored at 4 ◦C in a refrigerator while not in use. The stocksolution was diluted four folds daily as the working solu-tion (R1) before using. The other chemicals were used aspurchased without further purification. 0.5 g l−1 CTAB solu-tion (R2) was prepared by dissolving appropriate amount ofCeotl21tsowpasssfiNm[taplatwObfo

etection limit was 0.03 �g P l−1 (1 nmol l−1) for 0.18 ml sam-le [16]. Both VMoP and MoP could act with luminal toroduce CL light. The analysis of orthophosphate in seawa-er using the CL technique was not successful because ofhe seawater matrix interference. Zui and Birks [12] foundhat more than 150 mg of Ca2+ or 170 mg of Mg2+ per litreould affect the determination of 0.032 �mol l−1 orthophos-hate. Therefore, to determine the orthophosphate in seawatersing CL technique, it is very important to eliminate the matrixnterference.

Solid-phase extraction (SPE) technique is well known toave high enrichment factors and low reagent consumption17], and thus widely applied for analyte preconcentration andatrix elimination. Among various sorbent materials C18 ispopular one to be used for preconcentration of non-polar

rganic compounds and clean-up purposes. The previous workad found that MoP could interact directly with cationic sur-actant, cetyltrimethylammonium bromide (CTAB), to producehe MoP–CTAB ion-pair complex [18]. In this work, it wasonfirmed that the MoP–CTAB ion-pair complex could be effi-iently extracted by C18 cartridge and separated from the matrixf seawater, and a novel method was thus established based onhe oxidation of luminol by MoP–CTAB eluted from C18 car-ridge to avoid the interferences of Ca2+ and Mg2+ to the CLetermination.

To establish a CL method available for the rapid determina-ion of ultratrace orthophosphate in seawater, a flow injectionFI) coupled with luminescence analyzer was adopted to ensurehe automatically and reproducible in-line reagent solution mix-ng, MoP–CTAB enrichment on C18, MoP–CTAB eluting from18 and reacting with luminol, as well as CL determination.

TAB in water. The eluent was 0.3 mol l−1 sulphuric acidthanol solution made by dissolving the appropriate amountf concentrated sulfuric acid in ethanol. Luminol stock solu-ion (10.0 mmol l−1) was prepared by dissolving 0.178 g ofuminol (5-amino-2,3-dihydro-1,4-phthalazinedione, Fluka) in0 ml of carbonate buffer (0.1 mol l−1, pH 10.5), made up to00 ml with water and stored at 4 ◦C. A working luminol solu-ion (2 mmol l−1) was prepared by diluting 100 ml of the stockolution to 500 ml borate buffer prepared by dissolving 19 gf Na2B4O7·10H2O (GR) and 12 g NaOH of (GR) in 500 mlater. Orthophosphate stock solution (9.687 mmol l−1) was pre-ared from KH2PO4 dried at 105 ◦C for 2 h, and was storedt 4 ◦C while not in use. Orthophosphate working standardolution (4.843 �mol l−1) was prepared daily by diluting thetock solution with an appropriate amount of water. Silicatetock solution (10 mmol l−1) was prepared from Na2SiF6. Arti-cial seawater (salinity 35) was obtained by dissolving 41.5 gaCl and 15 g MgSO4·7H2O in 1.5 litre water according to theethod described in Specifications of Oceanographic Survey

19]. Orthophosphate in the artificial seawater was removed byhe MAGIC method [11]. The addition of 1 mol l−1 NaOH tortificial seawater at a 1:40 v/v ratio could produce Mg(OH)2recipitate that effectively scavenged orthophosphate out of theiquid phase and onto Mg(OH)2 surface. The precipitate wasged and allowed to settle down overnight. The supernatant washen siphoned off as phosphate-free artificial seawater, whichas used as the matrix for preparing standard curve solutions.rthophosphate-free seawater was prepared by the removal ofackground orthophosphate in low-nutrient seawater collectedrom the South China Sea, following the same procedure as forrthophosphate-free artificial seawater.

186 Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190

Fig. 1. FI manifold: P1, P2, PMT, C and W represent pump 1, pump 2, photo-multiplier tube, coil and waste, respectively.

2.2. Apparatus

All flasks used in the experiments were soaked with3 mol l−1 HCl for 2 h, cleaned with pure water in an ultra-sonic bath for 0.5 h, and rinsed thoroughly with Milli-Qwater.

A FI analysis system was fabricated with the following parts:an FIA 3110 flow injection analysis processor (Beijing TitanInstruments Co., Beijing, China) including two six-port peri-staltic pumps and a eight-way rotary valve; a thermostated bathwater (Shanghai Medical Instrument Co., Shanghai, China); aSep-Pak C18 cartridge (Waters Associates, Milford, MA, USA);and a BPCL luminescence analyzer (Institute of Biophysics,Chinese Academy of Sciences, Beijing, China) equipped with aflow cell (Ø 20 mm × 2 mm) with a 314 mm2 transparent quartzwindow, as shown in Fig. 1.

2.3. Flow injection manifold and procedures

The arrangement of the FI system is illustrated in Fig. 1. Allthe pump tubing was of silicon-latex. The other tubing was madeof PTFE.

The FI program is shown in Table 1. During step 1, appropri-ate amounts of R1 and R2 were simultaneously added to a 200 mlsample solution that was exactly measured into an Erlenmeyerflask in advance. The concentrations of ammonium heptamolyb-date tetrahydrate, sulphuric acid and CTAB in the final samplesolution were 0.17 g l−1, 0.036 mol l−1 and 13 mg l−1, respec-tively. After all the reagents were added, the mixed solution waspumped through a Sep-Pak C18 cartridge, and the MoP–CTABcompound was extracted onto C18 (step 3). Then the cartridgewas washed with water temporarily stored in the coil and thenethanol (step 4). During the elution step (step 5), the adsorbedMoP–CTAB was rapidly eluted by the eluent of 0.3 mol l−1 sul-phuric acid ethanol solution, then mixed in the flow cell withthe chemiluminescence reagent. The light produced by the reac-tion of luminol and MoP in the flow cell was detected with theluminescence analyzer, and the detector output was recordedwith a computer as an elution curve. The total CL intensity,equivalent to the integration of CL signal, was used to quan-tify the concentration of orthophosphate in sample. To obtaingood reproducibility, the eluent, luminol and sample solutionwere incubated at 35 ◦C. The photomultiplier tube voltage ofta[vwost

3

3

ouDti

Table 1Flow injection program

Step Time (s) Valve position Pump 1 flow rate (ml min−1) Pump 2(ml min−

1 90 Inject 3.5 (R1, R2); 6.5 (H2O) 02 50 Inject 0 233 300 Fill 0 234 70 Fill 6.5 (ethanol) 05 100 Inject 4.8 (eluent) 9.0 (luminol)6

20 Fill 6.5 (H2O, ethanol)

he detector was performed at −600 V. It should be pointed thatnalytical quality control procedures according to the Berger’s22] had been performed through out the experiments to pro-ide the sound data. For example, the blank and artificial sea-ater (salinity 35) sample solution containing 0.097 �mol l−1

rthophosphate were used as quality control samples. Thoseamples were measured with each set of 10 samples, to con-rol the detection variation in the designed range.

. Results and discussion

.1. Matrix elimination

The matrix of seawater would interfere in the determinationf orthophosphate with the CL technique. C18 cartridge wassed for MoP–CTAB preconcentration and matrix elimination.uring preconcentration, other impurities remained on the car-

ridge together with MoP–CTAB. Therefore, the washing stepn the time interval between the preconcentration and elution

flow rate1)

Comments

Adding reagents into sample, storing water in the coilPre-fillLoading sample solution onto Sep-Pak C18 (preconcentration)Water and ethanol washingElution, luminol mixed with eluent solutionRinsing Sep-Pak C18 with water

Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190 187

was found to be very important. The cartridge was rinsed with6 ml water then 1.6 ml ethanol to minimize matrix interferenceand to obtain a smooth elution curve. On the other hand, verysharp and abnormal elution curve would appear with only waterrinsing.

3.2. Parameter optimization

In order to establish optimal conditions for the lowest possibledetection limit in seawater matrix, the effect of various param-eters was investigated with a univariate experimental design.These were the composing of borate buffer; luminol, colordeveloping reagent and CTAB concentrations; eluent acidity;stopped flow time; preconcentration and eluting flow rate; effectof salinity and so on. All of these studies were performed withan artificial seawater (salinity 35) sample solution containing0.097 �mol l−1 orthophosphate.

3.2.1. The composing of borate buffer and theconcentration of luminol

CL of luminol is particularly dependent on the reaction pH[20]. The CL intensity was higher with borate buffer than withcarbonate buffer, and the maximum CL emission was observedat pH 12.5 [16]. Accordingly, borate buffer was chosen as thematrix of luminal solution in this work. Because the eluentsolution consisted of 0.3 mol l−1 sulphuric acid, it would neu-t(ttt0

o(lsfaa

3c

smff[www4raaw

Fig. 2. Variation of CL intensity with: (a) concentration of NaOH in boratebuffer; (b) concentration of luminol; (c) volume of R1 added in the final solution;(d) concentration of R2 in the final solution; (e) concentration of sulfuric acidin eluent.

of ammonium heptamolybdate tetrahydrate and sulphuric acidin the final sample solution were 0.17 g l−1 and 0.036 mol l−1,respectively.

CTAB (R2) interacted directly with MoP to produceMoP–CTAB ion-pair complex which could be efficientlyextracted by C18 cartridge. The effect of R2 on total CL inten-sity was shown in Fig. 2(d). The CL response increased as R2

ralize some OH− ion. The amount of NaOH in borate buffer0.1 mol l−1) was further investigated. As shown in Fig. 2(a),he total CL intensity was rapidly enhanced as NaOH increasedo 0.5 mol l−1 but no appreciable increase was observed abovehis concentration. Therefore, the borate buffer consisting of.1 mol l−1 borate and 0.6 mol l−1 NaOH was used for this study.

The effect of luminol concentration on the determinationf orthophosphate was studied over the range 0.1–4 mmol l−1

Fig. 2(b)). The total CL intensity was about constant when theuminol concentration was between 1.5 and 4.0 mmol l−1. Con-equently, a luminol concentration of 2.0 mmol l−1 was chosenor all subsequent experiments. The CL response varied with thege of the luminol solution and therefore luminol solution waslways prepared within 24 h before use [16].

.2.2. Color developing reagent (R1) and CTAB (R2)oncentrations

The amount of color developing reagent (R1) added in theample solution would affect on the MoP and MoP–CTAB for-ation. [H+]/[MoO4

2−] ratio was found to be a very importantactor affecting the color formation of PMB, and optimum colorormation occurred at [H+]/[MoO4

2−] ratio between 50 and 805]. Based on this, [H+]/[MoO4

2−] ratio was fixed on 74 in thisork, and the volume of R1 added in 200 ml sample solutionas studied. Illuminated as Fig. 2(c), maximum CL emissionas observed in the range of 4.0–5.0 ml. If R1 was less than.0 ml, the total CL intensity was decreased due to insufficienteagent for maximum formation of MoP. R1 above 5.0 ml wouldlso result in lower CL emission, because high acidity wouldffect the formation of MoP–CTAB. Consequently, 5.0 ml R1as added in 200 ml of sample solution, i.e., the concentrations

188 Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190

concentration increasing up to 13 mg l−1. Higher concentrationof R2 did not result in observable increase in signal. Therefore,13 mg l−1 R2 in the final sample solution was considered to beoptimal. In addition, Fig. 2(d) illuminates that the total CL inten-sity was very weak when R2 was absent. This hinted that MoPcould almost not be retained by C18 when R2 was absent dueto its hydrophilicity.

3.2.3. Eluent acidityIn our previous work, it was found that the ethanol solu-

tion consisting of 0.56 mol l−1 sulphuric acid could rapidlyelute PMB-CTAB ion-pair complex off the C18 cartridge [21].Accordingly, the ethanol solution consisting of sulphuric acidwas also used as the eluent in this study. The eluent acid-ity, expressed as the concentration of sulfuric acid in ethanolsolution, was further examined in the range between 0 and0.5 mol l−1. Fig. 2(e) confirmed that the MoP–CTAB adsorbedon C18 could be rapidly eluted by the ethanol solution consistingof more than 0.2 mol l−1 sulphuric acid, while pure ethanol couldnot completely elute the MoP–CTAB. In this work, the ethanolsolution consisting of 0.3 mol l−1 sulphuric acid was used.

3.2.4. Sample loading flow rateThe effect of sample loading flow rate was investigated in

the range of 19.0–31.0 ml min−1. The total CL intensity wasaol

3

ecCffl

3

iocwpIant

3

saiad

Fig. 3. The eluting curves obtained with different eluting flow rates: (A)5.1 ml min−1; (B) 4.8 ml min−1; (C) 4.3 ml min−1; (D) 3.5 ml min−1.

the effect of salinity, several different salinity artificial seawa-ter samples (salinity range 0–45) containing 0.097 �mol l−1

orthophosphate were analyzed following the proposed analyt-ical procedures. The results confirmed that the salinity had anegative effect on the formation of MoP–CTAB reflected by thetotal CL intensity. It is known that the salinity of open oceanwater is almost constant as 35. Therefore, artificial seawaterwith salinity 35 was chosen as the matrix for preparing standardcurve.

3.3. Interference of silicate

During the preconcentration step of MoP–CTAB, the anionsand cations of low molecular weight in seawater were notretained on C18 cartridge, and thus separated from MoP toCTAB. In the washing step the remained ions would be fur-ther washed away. Therefore, anions and cations would notdirectly interfere with the CL signal, but showing the influence asionic strength. The effect of inonic strength has been discussedabove (Section 3.2.7). On the other hand, Si will possibly formmolybdosilicic heteropoly acid with ammonium molybdate inthe proposed conditions, which would also be retained on thecartridge and have similar reaction as MoP with luminol, thusaffect the detection of orthophosphate. Therefore, the interfer-ence study was focused on Si.

tmoS2Top0idt

bout constant within the tested flow rate range. By balancingf column pressure, analysis time and reproducibility, a sampleoading flow rate of 23.0 ml min−1 was chosen.

.2.5. Time for formation of MoP–CTABIn the FI procedures, the stopped flow technique is sometimes

mployed to improve the sensitivity of FI method where thehemical reaction takes a longer time to complete. However, theL response did not be affected by the stopped flow time due to

ast formation of MoP–CTAB. It was concluded that the stoppedow technique was unnecessary in the proposed FI method.

.2.6. Eluting flow rateThe eluting flow rate was another important factor affect-

ng the elution of the MoP–CTAB. The eluting flow rate wasptimized from viewpoint of complete elution, peak shape andolumn pressure. The total CL intensity kept almost constantithin the tested flow rate range of 3.5–5.1 ml min−1, and theeak shortened and widened as the flow rate decreased (Fig. 3).n the experiment, the optimum eluting flow rate was chosens 4.8 ml min−1, where the elution was complete, the peak wasarrower as shown in Fig. 3 and the column pressure was notoo high.

.2.7. Effect of salinityHigher salinity corresponded to larger ionic strength in the

ample solution. According to Debye–Huckel equation, thectivity coefficients for ions in sample solution decreases asonic strength increasing. In other words, the activity of MoPnd CTAB ions, which is direct to the formation of MoP–CTAB,ecreases as salinity of the seawater increasing. To investigate

In order to investigate the effect of silicate present in seawa-er samples on orthophosphate determination using the proposed

ethod, artificial seawater samples containing 0.012 �mol l−1

rthophosphate and various amounts of silicate were analyzed.ilicate concentrations in these samples were between 0 and00 �mol l−1, which covers the range for most of natural waters.he results were tabulated in Table 3. The total CL intensityf 0.012 �mol l−1 orthophosphate in artificial seawater sam-les showed no significant variation within Si concentration of–200 �mol l−1. This indicated that even a 16,000-folds of sil-cate concentration in the sample solution did not influence theetection of 0.012 �mol l−1 orthophosphate compound. Thushe interference of silicate could be neglected.

Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190 189

Table 2Comparison of analytical results of the proposed method and MAGIC method

Seawater sample CPO43− ± S.D. (�mol l−1) Calculated t-value Critical t-value (P = 0.05)

Proposed method MAGIC method

1 0.060 ± 0.004 (n = 4) 0.058 ± 0.002 (n = 3) 0.850 2.5712 0.042 ± 0.005 (n = 4) 0.043 ± 0.004 (n = 4) 0.311 2.4473 0.022 ± 0.002 (n = 3) 0.025 ± 0.002 (n = 3) 1.612 2.776

3.4. Linearity, reproducibility and method detection limit

A good linearity between the total CL intensity andorthophosphate concentration was obtained within the con-centrations 0.005–0.194 �mol l−1. The linear equation forthe standard curve (with r = 0.9997 and P < 0.0001) wasTI = 6.13 × 107 C (�mol l−1) + 3.84 × 105 (n = 8), where TI isthe total CL intensity and C is the concentration of orthophos-phate in the artificial seawater sample.

The reproducibility of the method was evaluated by sevenrepetitive determinations of an artificial seawater sample witha concentration of 0.097 �mol l−1 orthophosphate over 8 hin 1 day. The mean value of the total CL intensity was6.18 × 106 ± 0.29 × 106 with R.S.D. 4.66%. The proposedmethod was proven to have reproducible analytical results. Itwas found that the analyses of at least 100 artificial seawatersamples could be done without replacing the cartridge.

The limit of detection (LOD) was calculated in accordancewith Berger’s [22]. Seven aliquots of the orthophosphate-freeseawater sample spiked with 0.0048 �mol l−1 orthophosphatewere analyzed following the proposed analytical procedures.The average result of the sample was 0.0048 �mol l−1 and thestandard deviation (S.D.) was 0.0007 �mol l−1. LOD was cal-culated using the following equation:

LOD = S.D. × t0.02,6 = 0.002 �mol l−1

Table 3Interference study of Si in the determination of orthophosphate in seawaters

CP (�mol l−1) CSi added (�mol l−1) Total CL intensity

0.012 0 1.01 × 106

50 1.01 × 106

200 1.03 × 106

3.5.2. Comparison with the MAGIC methodThree typical seawater samples, obtained from the South

China Sea, were analyzed using both the proposed methodand the MAGIC method [11]. The results are compared inTable 2. Using the paired Student’s t-test at 95% confidencelevel (P = 0.05) to test the difference between the two meth-ods, the calculated t-values shown in Table 2 were lower thanthe corresponding critical t-value. This indicates that there wasno statistically significant difference between the proposed andMAGIC methods (Table 3).

4. Conclusions

A novel on-line CL method was established to determineultratrace orthophosphate in seawater samples. The matrix inter-ference of seawater was successfully eliminated using C18solid-phase extraction. There was no statistically significant dif-ference between the results obtained from the proposed andMAGIC methods. Compared to the MAGIC method, the pro-posed method had the advantages of being more sensitive, faster,silicate interference free, sample saving and should have thepotential to be adopted in the field. The proposed method wassuitable for the open-ocean waters where the salinity was almostconstant as 35.

A

S

R

where t0.02,6, the two-tailed t-statistic at the 98% confidencelevel for six degrees of freedom, was 3.143. This LOD was lowenough to justify the use of this procedure for the determinationof orthophosphate in the South China Sea.

3.5. Validation of the method

3.5.1. RecoveryIn order to examine the method recovery, a series of

orthophosphate-free seawater samples spiked with orthophos-phate at 0, 0.024, 0.048, 0.097 and 0.194 (mol l−1 concentrationwere analyzed using the proposed method operated under theoptimized parameters. The linear equation for the matrix spikecurve (with r = 0.9989 and P < 0.0001) was TI = 5.67 × 107 CP(spiked, �mol l−1) + 4.02 × 105 (n = 5). The average recovery ofthe orthophosphate in spiked samples, represented as the ratio ofthe slope of standard curves prepared in seawater to that in arti-ficial seawater samples, was 92.5%. This high overall recoveryshowed that the real seawater matrix did not interfere with thedetermination of orthophosphate. The formation of MoP–CTABcompound, its retention on the C18 cartridge, and the elution ofthe compound from C18 seemed to be completed.

cknowledgment

The work was financially supported by the National Naturalcience Foundation of China (No. 40521003).

eferences

[1] S.V. Smith, Limnol. Oceanogr. 29 (1984) 1149.[2] M.D. Krom, N. Kress, S. Brenner, Limnol. Oceanogr. 36 (1991) 424.[3] S.L. Young, T. Seiki, T. Mukai, K. Takimoto, M. Okada, Water Res. 30

(1996) 1490.[4] J. Murphy, J.P. Rilly, Anal. Chim. Acta 27 (1962) 31.[5] L. Drummond, W. Mather, Anal. Chim. Acta 302 (1995) 69.[6] Y.L.L. Chen, H.-Y. Chen, D.M. Karl, M. Takahashi, Cont. Shelf Res.

24 (2004) 527.[7] A. Yimaz, S. Tugrul, J. Mar. Syst. 16 (1998) 253.

190 Y. Liang et al. / Analytica Chimica Acta 571 (2006) 184–190

[8] D.M. Karl, G. Tien, Mar. Chem. 56 (1997) 77.[9] K.K. Cavender-Bares, D.M. Karl, S.W. Chisholm, Deep Sea Res. Part

I: Oceanogr. Res. Papers 48 (2001) 2373.[10] M. Shoji, W. Toshiaki, T. Kyoji, Talanta 31 (1984) 235.[11] D.M. Karl, G. Tien, Limnol. Oceanogr. 37 (1992) 105.[12] O.V. Zui, J.W. Birks, Anal. Chem. 72 (2000) 1699.[13] J.-Z. Zhang, J. Chi, Environ. Sci. Technol. 36 (2002) 1048.[14] J.L. Haberer, J.A. Brandes, Mar. Chem. 82 (2003) 185.[15] I.P.A. Morais, M. Miro, M. Manera, J.M. Estela, V. Cerda, M.R.S. Souto,

A.O.S.S. Rangel, Anal. Chim. Acta 506 (2004) 17.

[16] M. Yaqoob, A. Nabi, P.J. Worsfold, Anal. Chim. Acta 510 (2004) 213.[17] L.A. Berrueta, B. Gallo, F. Vicente, Chromatographia 40 (1995) 474.[18] C.Z. Katsaounos, D.L. Giokas, A.G. Vlessidis, E.K. Karayannis, M.I.

Paleologos, Sci. Total Environ. 305 (2003) 157–167.[19] State Oceanic Administration of China, Specifications of Oceanographic

Survey, Ocean Press, Beijing, 1991, p. 179.[20] K. Robards, P.J. Worsfold, Anal. Chim. Acta 266 (1992) 147.[21] Y. Liang, D.X. Yuan, Q.M. Lin, Chin. J. Anal. Chem. 33 (2005) 1053.[22] W. Berger, H. McCarty, R.K. Smith, Environmental Laboratory Data

Evaluation, Genium Publishing Corporation, GA, USA, 1996, pp. 2–11.