Analysis for Cd, Cu, Ni, Zn, and Mn in estuarine water by ...kbruland/Manuscripts/BRULAND/Beck... · is highly sensitive and allows for the simultaneous analysis of numerous elements

Analytica Chimica Acta 455 (2002) 11–22

Analysis for Cd, Cu, Ni, Zn, and Mn in estuarine water byinductively coupled plasma mass spectrometry coupled

with an automated flow injection system

Nicole G. Becka, Robert P. Franksb, Kenneth W. Brulandc,∗a Department of Earth Sciences, University of California at Santa Cruz, Santa Cruz, CA 95064, USAb Institute of Marine Sciences, University of California at Santa Cruz, Santa Cruz, CA 95064, USA

c Department of Ocean Sciences, University of California at Santa Cruz, Santa Cruz, CA 95064, USA

Received 26 June 2001; received in revised form 15 November 2001; accepted 23 November 2001

Abstract

A flow injection inductively coupled plasma magnetic sector mass spectrometry (FI-ICP-MS) method was developed forthe analysis of Cd, Cu, Ni, Zn, and Mn in estuarine waters. The method uses just greater than 3 ml sample, and employs anautomated on-line preconcentration step using a metal chelating resin (Toyopearl AF-Chelate 650 M). Acidified samples forthe analysis of Cd, Cu, Ni, and Zn were buffered on-line to pH of 5.6 ± 0.2 with ammonia acetate just prior to loading ontothe chelating resin, while samples for Mn analysis were adjusted to pH of 9.0±0.2 prior to concentration. Limits of detectionwere: Cd= 1.4 ng l−1 (0.013 nM), Cu= 17 ng l−1 (0.27 nM), Ni = 28 ng l−1 (0.48 nM), Zn= 46 ng l−1 (0.70 nM), andMn = 86 ng l−1 (1.6 nM). The blank concentrations were less than 1.5% of the SLEW-2 concentrations for each elementanalyzed, except Ni, which had a significant, but very constant blank from the Ni cones used. The detection limits were lessthan 5% of the concentrations observed in the San Francisco Bay estuarine samples, with the exception of Zn where thedetection limit was 10% of the concentration of lowest San Francisco Bay sample analyzed. Using the FI preconcentrationtechnique, we conducted medium resolution scans of potential isobaric interferences (mass units 55–66) using a magneticsector ICP-MS to identify a number of interferent complications with evaluating certain isotopes, especially,59Co, and60Ni.Investigations of potential interferents illustrate the importance of appropriate isotope selection and, in some instances, theneed to perform blank corrections with both an instrumental and a matrix blank. The method was verified by the analysis ofan estuarine water standard reference materials (SLEW-2), and San Francisco Bay samples with previously reported values.Six estuarine samples run in triplicate generated the following average precision (presented as % R.S.D.); Cd= 4.2%,Cu = 3.2%, Ni = 3.3%, Zn= 4.4%, and Mn= 2.2%. © 2002 Published by Elsevier Science B.V.

Keywords: Trace metal analysis; Seawater; ICP-MS; Flow injection

1. Introduction

Inductively coupled plasma high resolution massspectrometry (ICP-MS) is a powerful new analytical

∗ Corresponding author.E-mail address: [email protected] (K.W. Bruland).

tool for the determination of trace metals, such asCd, Cu, Ni, Zn, and Mn in natural waters. ICP-MSis highly sensitive and allows for the simultaneousanalysis of numerous elements and their respectiveisotopes. The elevated levels of Mg2+, Ca2+, Na+,and K+ in seawater ([Mg]= 53 mM, [Ca]= 10 mM,[Na] = 468 mM, [K] = 10 mM) can cause a variety

0003-2670/02/$ – see front matter © 2002 Published by Elsevier Science B.V.PII: S0003-2670(01)01561-6

12 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

of problems and make direct analysis of trace metalsin seawater difficult, if not impossible, by ICP-MS.For proper operation, total dissolved solids (TDS)in samples injected into an ICP-MS should remainbelow 1000 mg l−1, presenting obvious problemswith seawater where TDS levels are on the order of35,000 mg l−1. The high salt concentrations can causeinstrument instability due to clogging of the nebulizer,injector, and/or cones. In addition, potential interfer-ences can be due to the formation of oxide and argidecomplexes such as CaO, MgAr, and NaAr. For thesereasons, separation and concentration of trace metalsfrom the bulk of the major ion salt matrix of seawaterprior to ICP-MS analyses is advised.

Traditional concentration techniques for these tracemetals include liquid/liquid organic extraction using acombination of 1-pyrrolidinedithiocarbamate/diethyl-dithiocarbamate (PDC/DDC) to simultaneouslyconcentrate the trace metals and remove the saltmatrix interference [1]. The PDC/DDC organicsolvent extraction method, while very reliable, istime consuming. Wells and Bruland [2] developeda modification of this solvent extraction methodinvolving a liquid/solid phase extraction of dithio-carbamate metal chelates, in this case with themore water-soluble dihydroxyethyldithiocarbamateas the ligand. Other investigators developed a metalchelate precipitation technique using PDC (1-pyrro-lidinedithiocarbamate/diethyldithiocarbamate) andcobalt for the determination of a variety of trace met-als in seawater [3]. Wu and Boyle [4] developed amethod involving coprecipitation of trace metals withMg(OH)2 (by simply raising the pH of seawater)and isotope dilution ICP-MS. These latter methodsrequire less chemistry than the solvent extractiontechniques, but are still more time consuming thanon-line preconcentration methods.

In the last decade, researchers have developedreliable methods of on-line preconcentration and ma-trix removal for trace metal analysis in seawater bycoupling flow injection (FI) analysis with ICP-MS[5–10]. Warnken et al. [10] provided a slightly dif-ferent on-line method using Toyopearl AF-Chelate650 M, showing that an ammonium chloride bufferand variations in the elution acids lowered detectionlimits and improved sensitivity for seawater tracemetal analyses. The method by Warnken et al. [10]further illustrates the potential for the adoption of

direct seawater analyses by FI-ICP-MS as an efficientand reliable analytical technique.

The purpose of this study was to develop a rapidmethod to analyze estuarine samples of variable salin-ities for trace metals and investigate matrix-based in-terferences using an ICP-MS coupled with the FI sys-tem. The magnetic sector ICP-MS [11] allowed us toinvestigate potential isobaric interferences in mediumresolution (R = 3500) at the isotopes of interest inorder to ensure proper isotope selection for each ele-ment. Previous FI-ICP-MS techniques have utilizedquadrupole systems [9,10,12], which do not have thecapability to separate the polyatomic interferent sig-nal from that of the trace metal. We will illustrate therelative intensities expected as a result of the poly-atomic interferences for mass units 55–66, and raisesome concerns about previous seawater analyses usingFI-ICP-MS for low level trace metal determinations.

We used a Finnigan MAT�-Sampler with a CetacASX-100 autosampler in order to automatically ana-lyze a sequence containing 24 samples, 4 blanks, and4 standard solutions in triplicate. Buffered estuarinesamples were loaded onto a commercially availablechelating resin, Toyopearl AF-Chelate 650 M, whichsuccessfully sequesters trace elements. Following anammonia acetate rinse of the column, the trace met-als were eluted into the plasma with 1.5N nitric acid.The completely automated on-line preconcentrationmethod presented herein can reliably analyze estuarinesamples for trace metals over a wide range of salinities.

2. Experimental

2.1. Reagents

All solutions were prepared with de-ionized water(18 M� cm−1) from a Milli-Q analytical reagent gradewater purification system (Millipore, Bedford, MA).Highly purified HNO3 was prepared by a sub-boilingquartz distillation of trace metal grade nitric acid (pur-chased from Fisher Scientific). The purified HNO3was diluted with Milli-Q to make up 1.5N HNO3 elu-ent, and then spiked with 10�g l−1 103Rh to be usedas an internal standard. Purified ammonia hydrox-ide was prepared by bubbling ammonia gas throughquartz sub-boiled distilled water. The 0.5 M ammo-nium/ammonium acetate buffer at pH= 5.6 ± 0.2

N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22 13

was prepared by diluting≈7.2 ml of trace metal gradeacetic acid (Fisher Scientific) and≈8.5 ml of puri-fied ammonia hydroxide in 250 ml of Milli-Q water.The pH= 9.0 buffer was prepared in a similar fash-ion by diluting≈7.0 ml of acetic acid and≈9.5 ml ofNH4OH in 250 ml of Milli-Q water. Stock solution(1000�g l−1) of Cu, Cd, Mn, Ni, and Zn were dilutedwith Milli-Q water to prepare standard solutions. Thesame rinse solution was used for all of the metal anal-yses and prepared by a 10-fold Milli-Q dilution of thepH = 5.6±0.5 M ammonium acetate buffer. All sam-ple handling was done with trace metal techniques [1]in a clean room, although the actual analyses wereperformed in a regular laboratory environment.

National Research Council of Canada (Ottowa,Ont., Canada) certified reference material of tracemetals in estuarine water (SLEW-2) with a salinity of11.6 was used to assess the accuracy of the method. Inaddition, accuracy was also evaluated by the analysisof filtered estuarine samples collected from the SanFrancisco Bay on 2 September 1995 at US GeologicSurvey water sampling stations BC10 and BC20. Theoriginal Cd, Cu, Ni, and Zn values from the San Fran-cisco Bay samples were generated by the laboratoryof A.R. Flegal by the PDC/DDC organic extractionmethod, and subsequently analyzed by graphite fur-nace atomic absorption spectroscopy (GFAAS) andpublished in the 1994 Regional Monitoring ProgramReport for Trace Substances [13]. Manganese con-centrations were later determined on the same SFBay samples by differential pulse cathodic strippingvoltammetry (DPCSV) [14]. All of the estuarine sam-ples were acidified to a pH of 1.6 shortly followingcollection to preserve the trace metals in solution.

2.2. Instrumentation and FI plumbing

We used a Finnigan MAT Element magnetic sec-tor ICP-MS with a Glass Expansion Conikal nebu-lizer and a Glass Expansion Cinnibar spray chamber(cooled to 10◦C). Standard nickel cones (SpectronInc.) were used. The Ni in the cones resulted in aconsistent Ni blank of about 100 ppt. The use of Ptcones would significantly reduce this blank further.The instrument operating parameters and data ac-quisition details are reported in Tables 1 and 2. Thenebulizer gas flow was optimized daily to providemaximum counts for the103Rh in the eluting acid

Table 1Instrumentation and operating parameters

Instrument Finnigan MAT Elementmagnetic sector ICP-MS

Spray chamber Glass Expansion CinnibarNebulizer Glass Expansion ConikalIncident power 1250 WCones Nickel, Spectron Inc.Plasma gas flow 13 l min−1

Auxiliary gas flow 0.75 l min−1

Nebulizer gas flow 0.85–0.95 l min−1 (tuned daily)Nebulizer sample flow 1.0 ml min−1

Autosampler CETAC ASX-100

Column componentsChelating resin Toyopearl AF-Chelate-650 M

resin (Tosohass,Montgomeryville, PA)

Sample column Global FIA (MC-ICNM)

(average sensitivity was 250 MHz ppm−1). The outputfrom the Finnigan�-sampler was connected directlyto the ICP-MS flow chamber. The ICP-MS and FIsystem were both controlled by the ICP-MS Elementsoftware. Teflon tubing (0.01, 0.02 and 0.03 in. i.d.:Upchurch Scientific Inc., Oak Harbor, WA, USA)was used as transfer lines between peristaltic tubingother connections. Initially, variable zinc contamina-tion was detected in tygon peristaltic pump tubing.In order to significantly lower the zinc blank levels,silicone peristaltic pump tubing (Rainin InstrumentCo. Inc., Woburn, MA) was used for all pump tubes,except the buffer and rinse lines which were protectedby clean-up columns. Connections were made usingTefzel flangeless fittings (Upchurch Scientific, OakHarbor, WA, USA). Toyopearl AF-Chelate-650 Mresin (Tosohass, Mongomeryville, PA, USA) waspacked into a Global-FIA column (MC-1CNM). The1 cm column consisted of a tapered inner chamberwith frits at each end which are push-fit into a threadedouter sleeve. A Watson Marlow PumpPro MPL peri-staltic pump was used to pump samples from the autosampler to the nebulizer and to remove waste fromthe nebulizer. A schematic of the FI plumbing is pre-sented in Fig. 1. The i.d. of the pump tubing and thepump speeds were manipulated so that flow within allof the reagents and sample lines was maintained at1 ml min−1. Two additional 2 cm Global-FIA columnswere packed with the iminodiacetate resin and

14 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

Table 2Operating parameters and isotope information

Data acquisition

Isotope Mass window Settling time Sample time Samples/peak Detection mode Runs/passes

55Mn 5 0.100 0.0100 100 Analog 165/162Ni 5 0.100 0.0100 100 Counting 123/163Cu 5 0.001 0.0100 100 Analog 123/166Zn 5 0.001 0.0100 100 Counting 123/1111Cd 5 0.001 0.0200 100 Counting 123/1

Isotope Isotopic abundance (%) NH4Ac buffer pH Detection limit (3σ ) (ng l−1)

55Mn 100 9.0± 0.2 8662Ni 3.59 5.6± 0.2 2863Cu 69.2 5.6± 0.2 1766Zn 27.9 5.6± 0.2 46111Cd 12.8 5.6± 0.2 1.4

installed in the buffer and wash lines after the pumpto remove any metals in the buffer and rinse solutionsand decrease subsequent blanks and detection limits.

2.3. General procedure

The �-SAMPLER operation program is presentedin Table 3. The prefill step allows the new sample tofill and rinse the sample tubing and at the same timeconditions the column with rinse solution prior to load-ing the buffered sample (Step 1). The 1 ml of bufferedsample is then loaded onto the column (60 s), followedby a 30 s, 0.05 M ammonium acetate rinse at pH= 5.6(Step 2). In Step 3, the sample is eluted off the col-umn with 1.5N HNO3 into the plasma for 70 s in orderto see the signal and subsequently allow trace metalconcentrations to return close to background levels.

The autosampler holds 24 7 ml samples, four 15 mlstandard solutions, in addition to a 15 ml 1% HNO3sample tubing rinse. At a flow rate of 1 ml min−1, eachanalysis consumes a little more than 3.0 ml of sam-

Table 3Finnigan mat�-sampler flow injection program

Step Time (s) Pump 1 (rpm) Pump 2 (rpm) Valve Remarks

Prefill 30 0 16.0 A Condition column, prefill sample1 (load) 60 0 16.0 B Load column2 (rinse) 30 0 16.0 A Rinse column3 (elute) 70 32.0 0 A Elute and acquire data

ple. This configuration allows analysis of a sequenceof 24 samples, including triplicates of standards (twowere used routinely) and blanks for Cd, Cu, Ni, andZn using the 5.6 pH buffer. The buffer solution is thenchanged to pH= 9.0 and the same samples rerun in se-quence to obtain Mn concentrations. The total time fortwo runs of 33 analyses (24 samples) is just under 4 h.

2.4. Data analysis

Time series elution peaks were generated for eachsample for the trace metals evaluated (Fig. 2). Nosmoothing of the data was done. In order to determinesample concentrations, an average peak count wasdetermined over a 15 s time period during the elutionpeak. The exact boundaries for the time window weredetermined on a daily basis, due to small variationsin the flow rate caused by wear of the pump tubes. Abaseline value was determined by averaging over a 15 swindow near the end of 70 s elution period and thiswas subtracted from the average peak count value. All

N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22 15

Fig. 1. FI �-SAMPLER pumping configuration. (A) Prefill and/orrinse column position; (B) load column with buffered sampleposition; (C) elute trace metals from column with 1.5N HNO3.W: waste; C: reagent column; solid line: liquid flow; dotted line:no flow.

samples and standards counts were corrected for blankvalues measured during the analysis. Blank concen-trations were less than 1.5% of the SLEW-2 concen-trations for each of the elements investigated. Linearcalibration curves were constructed and sample metalconcentrations were calculated. The data analysis andstandard calibration process has been streamlined us-ing a Microsoft Excel visual basic macro. Rhodiumis not retained on the Toyopearl column, and there-fore could be used as an instrumental internal standardby spiking the elution acid with 10�g l−1 to correctfor variations in instrument sensitivity over the courseof an analytical run. Our experience with the Ele-ment ICP-MS is that103Rh works well as an internalstandard for the first row transition elements. In gen-eral, variations in the103Rh counts were small (<5%).When larger variations were observed, all count rateswere corrected appropriately by the103Rh values.

2.5. Results and discussion

We experimented with various pH levels for the am-monia acetate buffer to optimize the metal recoveriesand analytical precision. The buffer pH of 5.6 ± 0.2yielded the highest recoveries of Cd, Cu, Ni, and Zn,but the highest count rates for Mn were obtained witha pH of 9.0 ± 0.2 buffer (supporting the findings ofWillie et al. [9]). We make our best attempts to min-imize the potential from metal-organic complexation(with Cd, Cu, Ni, and Zn in the natural water sam-ple) by preserving the samples in an acidified solution(pH = 1.6) and buffering the sample on-line less than3 s before the buffered sample reaches the iminodiac-etate resin. While we are unable to quantify the lossof any trace metals from the sample to organic com-plexation in this short amount of time, the respectiverecovery of this method against certified values sug-gests that at estuarine levels that organic complexationmay not be cause for concern.

We varied the time spent rinsing the column with the0.05 M (pH= 5.6) ammonium acetate rinse solutionfrom 5 to 60 s to investigate the removal of Na, Mg,and Ca from the column. Preliminary results showedno decrease for Ca, a small initial decrease for Mg,and a significant decrease for Na. This is not surpris-ing, since both Ca and Mg have reasonable affinitiesfor the iminodiacetate resin. We chose to use a 30 srinse step since that removed the majority of the Na

16 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

Fig. 2. Elution peaks for SLEW-2. (A)55Mn and 103Rh elution following the buffered sample loading (0.5 M ammonium acetatepH = 9.0 ± 0.2); (B) 111Cd, 63Cu, 62Ni, 66Zn, and 103Rh elution following the buffered sample loading (0.5 M ammonium acetatepH = 5.6 ± 0.2). The column rinse for both methods was done with a 0.05 M ammonium acetate solution. In order to place the elutionpeaks on a similar scale55Mn, 63Cu, 103Rd counts are divided by a factor of 5, 10, and 50, respectively, and111Cd counts are increasedby an order of magnitude.

and minimized the analysis time. It is possible that us-ing a different rinse solution, or rinsing for a longerperiod of time might have further reduced the levelof interfering species. We also discovered a variableamount of zinc contamination when the acidic solu-tion would remain in the Tygon peristaltic pump tub-ing while the other pump was in operation. Changingto silicone tubing greatly reduced the Zn blank and im-proved analytical precision and the Zn detection lim-its. Whether we created standard and blank solutions

in an open-ocean seawater or a Milli-Q matrix, therewas no detectable difference in analytical accuracy ofthe SLEW-2 trace metal concentrations. Six estuarinesamples were run in triplicate to evaluate the analyt-ical precision for each element. The average relativestandard deviations are as follows: Cd= 4.2%, Cu=3.2%, Ni = 3.3%, Zn= 4.4%, and Mn= 2.2%.

Any variability in instrument performance as a re-sult of high salt content is greatly reduced by the sep-aration and concentration of the trace metals on the

N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22 17

chelating resin. The use of Rh as an internal standardin the elution acid allows the correction for any vari-ations in signal intensities due to plasma instabilityor nebulization variability across samples. However,there are two forms of potential polyatomic interfer-ences that are much more difficult to separate fromthe actual signal of the trace metal of interest. Theseinclude: (1) argon species inherent in the basic opera-tion of an argon plasma ICP-MS (i.e. ArO, ArN, andArOH), and (2) sample matrix interferences as a resultof high concentrations of other elements in the sample(such as Ca, Mg, K, and Na) resulting in the formationof metal oxide and metal argide species (i.e. MgAr,NaAr, KO, CaO and MoO). We used the medium res-olution capability (R = 3500) of the ICP-MS to in-

Table 4Potential interferences for isotopes of interesta

Mass Elements Isotopic abundance (%) Potential polyatomicinterference

Interference isotopicabundanceb (%)

55 55Mn 100 39K16O 9315N40Ar 0.439Ar16OH 0.06

56 56Fe 91.7 40Ar16O 9940Ca16O 97

57 57Fe 2.2 40Ar17O 0.0440Ar16OH 9941K16O 7

58 58Fe 0.28 40Ar18O 0.258Ni 68.3 40Ar17OH 0.04

42Ca16O 0.640Ca18O 0.2

59 59Co 100 40Ar18OH 0.243Ca16O 0.123Na36Ar 0.3

60 60Ni 26.1 44Ca16O 224Mg36Ar 0.2

61 61Ni 1.13 23Na38Ar 0.0662 62Ni 3.5963 64Cu 69.2 23Na40Ar 99.9

64 64Ni 0.91 48Ca16O 0.264Zn 48.6 24Mg40Ar 79

65 65Cu 30.8 25Mg40Ar 1066 66Zn 27.9 26Mg40Ar 11

111 111Cd 12.8 95Mo16O 16

114 114Cd 28.7 98Mo16O 24114Sn 0.65

a Identification of interferences and relative abundance were obtained directly from the Finnigan element 2 software; InterferencesWorkshop Version 2.3, developed by R. Pesch.

b Interference isotopic abundance is the product of the individual isotope abundances.

vestigate potential interferences that may contribute tothe final magnitude of an analytes’ intensity and re-sult in inaccurate concentrations, even following thepreconcentration step.

In medium resolution, we explored each of theisotopes for the elements of interest and determinedif molecular polyatomic species had any contributionto our analyte intensities. Although we only devel-oped methods to measure Cd, Cu, Ni, Zn and Mn inestuarine samples, we investigated all of the potentialinterferences applicable to the isotopes from mass55–66 (Table 4). Argon polyatomic species inherentin the ICP-MS operations (such as ArO and ArOH)will be present in all samples, standards, and blanks.Although the amounts of these species can vary

18 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

Table 5National Research Council of Canada SLEW-2 validation resultssalinity = 11.6,n = 5

Analyte Certified value(�g l−1)

Experimentalvalue (�g l−1)

Recovery (%)

Cd 0.019± 0.002 0.021± 0.007 110± 37Cu 1.62± 0.11 1.60± 0.04 99± 7Ni 0.709 ± 0.054 0.665± 0.051 94± 11Zn 1.10± 0.14 1.21± 0.12 110± 16Mn 17.1 ± 1.1 14.8± 0.8 87± 18

significantly with sample matrix, we found only smallvariations. As long as the contribution from the inter-fering plasma species is small compared to the signalfrom the analyte element, these can be corrected forby using a Milli-Q blank. In contrast, interferencesdue to the salt matrix are only present in the samples,vary with sample salinity, and cannot be corrected forwith a Milli-Q blank. Figs. 3 and 4 present the com-parison of an acidified Milli-Q blank (Figs. 3A and4A), an open-ocean seawater blank (1 l of seawater(salinity = 33) treated with 50 mg of Chelex-100 pe-riodically shaken vigorously and left for over 5 days)(Figs. 3B and 4B), a 10�g l−1 multi-element 1%HNO3 tune solution (Figs. 3C and 4C), and a SLEW-2sample (Figs. 3D and 4D). The certified values for theSLEW-2 trace metals are presented in Table 5. Figs. 3and 4 allow the identification of the instrumental in-terferences, such as ArO and ArOH species, versesthose species formed as a result of the salt contentof the samples. For example, the40Ar16OH signalat mass unit 57 is the same intensity in the Milli-Qblank, seawater blank, and tune solution (6% differ-ence of the maximum counts across all three samples).Although most of the interfering species could beidentified, one at mass 61 could not. Every possiblecombination of plasma and matrix elements that couldproduce interferences at mass 61 should have alsoresulted in interferences at other masses, which werenot seen. The identity of this interference remains amystery.

The instrumental interferences present at mass 56,57, and 58 can be corrected for by simply calculat-ing the Milli-Q blank concentrations of the ArO andArOH species at the respective isotopes, but the rel-ative intensity of the argon species will have a di-rect influence on subsequent detection limits for56Fe,

57Fe, 58Fe, and58Ni (in most circumstances the iso-baric interference of58Fe and58Ni makes that massof little value). However, the analysis of the monoiso-topic 59Co is limited by the presence of43Ca16O, in-herent in saline waters, at an equivalent concentrationof 0.04�g l−1 in the seawater blank (Fig. 3B). Therelative intensity of the43Ca16O species makes theanalysis of a sample with a Co concentration of only0.055�g l−1 (such as SLEW-3) questionable, at leastwith this separation scheme. This interferent cannotbe resolved by Milli-Q blank correction. The analysisof isotopes containing salt matrix polyatomic speciesis limited by the relative concentration of trace metalin the salt-matrix blank to that of the sample, makingestuarine samples with high levels of59Co or60Ni forexample, correctable, but determination of open-oceansamples by this method suspect. Similar polyatomicinterferents are present at mass units 61–66 (Fig. 4Band C) and pose similar problems when attempting toresolve the intensity due to the trace metal in a lowlevel sample (Fig. 4D).40Ca16O and40Ca16O1H mayalso be a problem at masses 56 and 57, however, theabundance of40Ar16O and40Ar16O1H in our systemwas high enough that no additional contribution couldbe seen. Many researchers analyzing Ni in seawaterby FI-ICP-MS methods evaluate60Ni, the most abun-dant of the Ni isotopes [9,10,12]. However, our in-vestigations illustrate that44Ca16O persists at levelsequivalent to 0.6�g l−1 of Ni following the on-linepreconcentration treatment of a seawater sample, atleast with this separation scheme. Unfortunately, theinability of a quadrupole ICP-MS to resolve these twospecies and the difficulty in obtaining a Ni-free saltwater blank implies that uncorrected Ni results per-formed on quadrupole instruments may be suspect, es-pecially at low levels of Ni. We chose instead to eval-uate62Ni. While the23Na2

16O interferent at mass 62still causes a small error in the analysis, the error issignificantly reduced from that at mass 60.

Most previous FI-ICP-MS methods have evaluated114Cd rather [8,9,12] than111Cd. Wu and Boyle [4]investigated the respective isobaric interferences forCd isotopic analysis by ICP-MS, and corrected forthe interference by monitoring the Pd, Sn, and Moconcentrations within the seawater samples and thencalculating expected compound interferent level at theisotope of interest. In order to avoid the significant114Sn interferent, our method evaluates111Cd. Both

N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22 19

Fig. 3. Isobaric interference investigation in medium resolution of mass units 55–60. (A) Milli-Q blank; (B) open-ocean seawater blank;(C) 10 mg l−1 tune solution; (D) SLEW-2.

20 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

Fig. 4. Isobaric interference investigation in medium resolution of mass units 61–66. (A) Milli-Q blank; (B) open-ocean seawater blank;(C) 10 mg l−1 standard tune solution in 1% HNO3; (D) SLEW-2.

N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22 21

Fig. 5. Elution peak of111Cd and 0.5% of95Mo concentrations= 95Mo16O from SLEW-2.

of the Cd isotopes suffer from a potential interferenceby MoO (Table 4). Unfortunately, an instrumentalresolution of >32,000 is required to resolve this in-terference. Direct aspiration of a concentrated Mostandard showed that MoO was present at a count rateof ≈0.5% under the operating conditions used. Wemonitored the elution of95Mo in our system to seeif it was causing a significant interference at111Cd(Mo is normally≈ 100× the concentration of Cd inseawater). Fig. 5 shows the elution profile for111Cd(calculated from direct95Mo measurement and as-suming 0.5% of the total95Mo16O) and111Cd from aSLEW-2 sample. The small amount of95Mo16O ob-served in the sample is relatively insignificant whencompared to the111Cd peak.

This method illustrates the ability to utilizeFI-ICP-MS with estuarine waters of variable salin-ity and a wide range of trace metal concentrations.The experimental results of five replicate analysesof SLEW-2 (salinity ≈ 11.6) by FI-ICP-MS arecompared to the certified values in Table 5. In ad-dition, values compared to previously analyzed SanFrancisco Bay water samples collected from beneath

the Golden Gate Bridge (Station BC20) and nearYerba Buena Island (Station BC10) are presented inTable 6. Our values agree well with previous valuesobtained by accepted analytical techniques. We onlyfocused our efforts on the analysis of the five metalspresented herein, but our investigations of the poten-tial interferences and the FI method development byother researchers suggest that this technique shouldbe applicable to the analysis of Co, Pb, Fe, and othermetals with a high affinity for this resin.

There are distinct benefits to magnetic sectorFI-ICP-MS during method development for the anal-ysis of trace metals in seawater. The ICP-MS inmedium resolution has the capability to identify natu-ral and matrix interferences and evaluate the effects ofeach interferent on the particular isotopes of interest.Our research illustrates the importance of address-ing and resolving the polyatomic interferents priorto finalizing trace metal values. In addition, we areable to investigate the direct effect variations in pHof the buffered sample or rinse times have on the iso-topes we are investigating. These capabilities are notavailable to researchers using quadrupole systems.

22 N.G. Beck et al. / Analytica Chimica Acta 455 (2002) 11–22

Table 6Analytical results for San Francisco Bay water samples collected during February 1995

Analyte FI-ICP-MS value (�g l−1) R.S.D. (%) Organic extraction GFAASvalue (�g l−1)a analyticalprecision≈10%

DP CSV value (�g l−1)b

USGS station BC20 (Golden Gate) salinity≈ 33 (n = 5)Cd 0.027± 0.002 4.3 0.027Cu 0.79± 0.02 2.0 1.00Ni 1.07 ± 0.03 3.1 1.10Zn 0.468± 0.005 1.1 0.384Mn 2.29 ± 0.08 5.2 1.76± 0.07

USGS station BC10 (Yerba Buena Island) salinity≈ 15c (n = 5)Cd 0.033± 0.002 5.0 0.031Cu 1.67± 0.12 7.0 1.89Ni 2.03 ± 0.08 3.8 2.09Zn 0.785± 0.061 7.8 0.896Mn 5.53 ± 0.09 1.8 5.12± 0.20

a [13].b [14].c [15].

Acknowledgements

Special appreciation is given to the analytical lab-oratory of Professor A.R. Flegal who provided uswith San Francisco Bay water samples previously an-alyzed for trace metals. This research was supportedby the Harbor Processes Program of ONR Grant no.N00014-99-1-0035.

References

[1] K.W. Bruland, R.P. Franks, G.A. Knauer, J.H. Martin, Anal.Chim. Acta 105 (1979) 223–245.

[2] M.L. Wells, K.W. Bruland, Marine Chem. 63 (1998) 145–153.[3] E.A. Boyle, J.M. Edmond, Determination of trace metals

in aqueous solution by APDC chelate co-precipitation, in:T.R.P.J. Gibbs (Ed.), Analytical Methods in Oceanography,1975, American Chemical Society, Washington, DC.

[4] J. Wu, E.A. Boyle, Anal. Chem. 69 (1997) 2464–2470.[5] E. Heithmar, T. Hinners, J. Rowan, J. Riviello, Anal. Chem.

62 (1990) 857–864.

[6] M.J. Bloxham, S.J. Hill, P.J. Worsfold, J. Anal. At. Spectrom9 (1994) 935–938.

[7] J.W. McLaren, J.W.H. Lam, S.S. Berman, K. Akatsuka, M.A.Azeredo, J. Anal. At. Spectrom. 8 (1993) 279–286.

[8] R.A. Nickson, S.J. Hill, P.J. Worsfold, Anal. Chim. Acta 351(1998) 311–317.

[9] S.N. Willie, Y. Ilda, J.W. McLaren, At. Spectrosc. 19 (1998)67–72.

[10] K.W. Warnken, D. Tang, G.A. Gill, P.H. Santschi, Anal. Chim.Acta 423 (2000) 265–276.

[11] U. Giessmann, U. Greb, Fresenius’ J. Anal. Chem. 350 (1994)186–193.

[12] H. Kumagai, M. Yamanaka, T. Sakai, T. Yokoyama, T.M.Suzuki, T. Suzuki, J. Anal. At. Spectrom. 13 (1998) 579–582.

[13] SFEI, Regional Monitoring Program Report for TraceSubstances in the San Francisco Bay, 1994.

[14] J.S. Roitz, K.W. Bruland, Anal. Chim. Acta 140 (1997) 99–113.

[15] J.L. Edmunds, B.E. Cole, J.E. Cloern, R.G. Dufford, Studiesof the San Francisco Bay, California, Estuarine Ecosystem,Pilot Regional Monitoring Program Results, 1995, USGeological Survey Open-File Report, Vol. 380, p. 97-15;US Geological Survey Open-File Report, Vol. 380, 1997,p. 97-15.