Sensitive determination of three arsenic species in water by ion exclusion chromatography-hydride generation-inductively coupled plasma mass spectrometry Tadashi Taniguchi,a Hiroaki Tao,*b Mamoru Tominagab and Akira Miyazakib aShimadzu Co. Ltd. 380–1 Horiyamashita Hadano Kanagawa 259–1304 Japan bNational Institute for Resources and Environment 16–3 Onogawa Tsukuba Ibaraki 305–8569 Japan. E-mail hiroaki@nire.go.jp Received 8th December 1998 Accepted 15th February 1999 A sensitive and robust speciation method for arsenic in water is described. The separation of arsenate (AsV) arsenite (AsIII) and monomethylarsonic acid (MMA) was performed using an ion exclusion column packed with a sulfonated polystyrene resin and using dilute trifluoroacetic acid at pH 2.1 as the mobile phase.The hydride generation method was used to improve sensitivity and to eliminate interference from chloride ions. The analysis time per sample was 18 min but could be shortened to 9 min by using a column switching method. The detection limits for the arsenic species were 1.1 pg ml-1 for AsV 0.5 pg ml-1 for AsIII and 0.5 pg ml-1 for MMA with an injection volume of 50 ml. The relative standard deviations of five replicates of a standard containing 1 ng ml-1 As of each species ranged from 0.8 to 2.8%. The method was validated by analyzing reference water samples. Introduction Since the toxicity of an element depends on its chemical form it is important to determine the concentration of individual chemical species in order to evaluate environmental risk.For arsenic and derivatives thereof the toxicity decreases in the order arsenite (AsIII )>arsenate (AsV )>[dimethylarsinic acid (DMA) monomethylarsonic acid (MMA)]>[arsenobetaine (AsB) arsenocholine (AsC) tetramethylarsonium ion (TMA)]. As a result the biological methylation of arsenic is generally thought to represent a detoxification process. However reports have appeared concerning the toxicity of DMA which includes DNA modification and mutagenicity.1,2 Arsenic in water occurs mainly in inorganic forms such as AsV and AsIII but also occurs in methylated forms such as MMA and DMA at very low concentration levels.3,4 The levels of other organoarsenic species found in biological tissues such as AsB AsC TMA and arsenosugars are negligible in water.A variety of methods for the speciation of arsenic have been developed thus far,5–8 and hyphenated methods such as liquid chromatography-inductively coupled plasma mass spectrometry (LC-ICP-MS) and capillary electrophoresis (CE)- ICP-MS,9 are currently popular and in common use. However most of these methods are not applicable to the direct determination of MMA and DMA in water because these species are present at low concentrations. It has also been diYcult to apply these methods to sea-water samples because of the high level of polyatomic interference from 40Ar35Cl on 75As. The purpose of the present investigation was to develop a sensitive and robust method for the determination of these species in water. As the analytes four arsenic species viz.AsIII AsV MMA and DMA were investigated because of their high level of toxicity. Another reason for selecting these species is that they form volatile hydrides. It is also possible to generate volatile hydrides from non-hydride forming species such as AsB and AsC provided that they are decomposed into inorganic forms by microwave digestion10,11 or photooxidation. 12,13 However these techniques were not investigated in the present study. In order to increase the sensitivity a variety of sample introduction methods such as high-eYciency nebulizers,14 ultrasonic nebulizers,15 thermospray nebulizers,16 direct injection nebulizers17 and hydride generation,3,4,9 have been used. Among these hydride generation (HG) resulted in the highest sensitivity for arsenic species.Furthermore since only gaseous species were introduced into the ICP via HG polyatomic ion spectral interference from 40Ar35Cl on 75As was not a factor. Clogging of the sampling cone and non-spectral interference such as space charge eVects were also eliminated. It is necessary not only to increase the sensitivity but also to decrease the background signal in order to improve the detection limit. The detection limits for arsenic with LC-ICP-MS are often determined by the high background signal arising from arsenic impurities in the LC eluent. In particular when a phosphate buVer4,10,16–19 was used as the eluent this phenomenon was pronounced because reagents derived from phosphorus which belongs to the same periodic group as arsenic generally contain significant arsenic impurities.Therefore it is desirable to perform a chromatographic separation of arsenic species via the use of only a high-purity acid solution as the eluent. The LC modes investigated thus far for arsenic speciation include reversed-phase,14 ion-pair reversed-phase,3,14,17–21 size exclusion,20 micellar22 and ion-exchange chromatography. 4,16,23–25 However in so far as we are aware there is no example in which the separation of arsenic species was performed based on ion exclusion. In the present study ion exclusion LC was examined for the separation of such species. Our data show that three arsenic species viz. AsIII AsV and MMA can be separated using only a dilute acid solution but that DMA cannot be eluted.The detection limits obtained with the present method were more than 20 times lower than those reported previously.3,4 This appears to be largely due to the use of dilute acid as the eluent which can easily be obtained in high purity compared with salts. Although deviations in retention times and background signal fluctuation 651 J. Anal. At. Spectrom. 1999 14 651–655 due to co-existing ions were reported for ion-pair reversedphase LC and ion-exchange LC,4,18 these interferences were not observed in the experiments reported herein suggesting that the method is of superior robustness. Because of this the present method is directly applicable to the analysis of seawater samples. 3 4 2O3 in a small portion Preliminary experiments revealed that a sea-water sample of Japan) and dimethylarsinic acid 3)2AsO(OH)] (Tri Chemical Laboratory) in water water containing 1 ng ml-1 of AsV was investigated.The As a result the eVect of adding NaCl or Na2SO4 to pure Results and discussion 4 4 3 3 Column Mobile phase Experimental 4 2 Reagents A stock solution of arsenate (AsV) at 1000 mg ml-1 As was prepared by dissolving sodium arsenate dibasic heptahydrate (Na HAsO ·7H2O) (reagent grade >99% Wako Osaka 1000 mg ml-1 As was purchased from Wako (atomic absorp- Japan) in water. A stock solution of arsenite (AsIII) at tion spectrometry grade). According to the manufacturer the EVect of Na2SO4 concentration on the AsV peak shape solution was prepared by dissolving As of NaOH solution which was then neutralized with HCl AsV gave a sharp peak but that AsV when added to pure solution to pH 5.0.Stock solutions of MMA and DMA at 1000 mg ml-1 As were prepared by dissolving monomethylarwater gave only a small and broad peak. It was speculated that some components contained in sea-water played an sonic acid [CH AsO(OH)2] (Tri Chemical Laboratory important role in the generation of the sharp peak for AsV. Yamanashi [(CH respectively. A 1% sodium tetrahydroborate solution was addition of NaCl had no eVect but added Na2SO4 produced prepared by dissolving high-purity grade NaBH4 (>95% a marked improvement in the peak shape although the reason Merck Darmstadt Germany or >98% atomic absorption for this was not apparent.The eVect of Na2SO4 concentration spectrometry grade Kanto Kagaku Tokyo Japan) in a 0.1 mol l-1 NaOH solution immediately prior to the experion AsV and AsIII peak shapes is shown in Fig. 2. The addition 2SO4 at 400 mg ml-1 as the final concentration was found of Na ment. Sodium hydroxide (analytical-reagent grade) and nitric to be suYcient and this concentration had no eVect on the acid (ultrapure grade) were purchased from Merck. equilibrium between AsV and AsIII. The added Na2SO4 did Trifluoroacetic acid (>99% protein sequencing grade) and not increase the blank level due to the presence of low levels sodium sulfate (anhydrous analytical-reagent grade) were of arsenic as impurities. purchased from Wako. Ultrapure water from a Milli-Q Low TOC system (Millipore Milford MA USA) was used EVect of NaBH concentration throughout.Instrumentation A schematic diagram of the LC-ICP-MS instrumentation with hydride generation is described in Fig. 1. An LC-6A liquid chromatograph (Shimadzu Kyoto Japan) equipped with a Shimadzu LC-6A pump and a sample injection valve (9725i Rheodyne Cotati CA USA) with injection volumes of 50 or 200 ml was used. The ion exclusion column was a sulfonated polystyrene type Shim-pack SCR-102H (30 cm long Shimadzu). A 5 cm guard column with the same packing material was also used in a column switching method. The mobile phase was dilute trifluoroacetic acid adjusted to pH 2.1 and a flow rate of 1.5 ml min-1 was used. The eluate from the liquid chromatograph was first mixed with 1.5 mol l-1 HNO3 (flow rate 2.3 ml min-1) and then with a 1% NaBH solution (flow rate 1.8 ml min-1) to generate the hydrides.The mixture was then transferred to a gas–liquid separator (fabricated in our laboratory) through a poly(tetrafluoroethylene) tube Fig. 1 Schematic diagram of the LC-ICP-MS instrumentation with hydride generation. 652 J. Anal. At. Spectrom. 1999 14 651–655 (50 cm long 3 mm id). The details of the gas–liquid separator have been described in a previous paper.26 The waste liquid was removed immediately after the reaction from the phase separator by means of a peristaltic pump to prevent the excess of hydrogen from entering the plasma and to prevent the chromatographic peak from broadening as a result of memory eVects.An ICPM-8500 inductively coupled plasma mass spectrometer (Shimadzu) equipped with a miniaturized torch was used. Operating conditions are listed in Table 1. The data from the ICP-MS instrument were converted into ASCII format and handled with an Excel spreadsheet (Microsoft Cambridge MA USA) for further processing. 4 7 1.2 1.5 5 20 75 0.64 1 1.8 1.5 2.3 4 The eVects of NaBH4 concentration on the signal peak heights and the background equivalent concentrations (BECs) for AsV AsIII and MMA are shown in Fig. 3. The BECs are defined as the concentrations that would give the equivalent peak height to the continuous background signal. Although the signal peak heights increased with increasing NaBH4 concentration the BECs became worse because of the increased background levels arising from impurities present in the NaBH4 solution.In order to obtain both considerable peak heights and better BECs 1% NaBH was chosen as a compromise concentration. Peak broadening by incorporating a hydride generation reaction between the LC and ICP-MS steps was negligible compared with the peak widths obtained by LC-ICP-MS without hydride generation. This is probably Table 1 Optimum operating conditions for LC-HG-ICP-MS ICP-MS parameters— Forward power/kW Ar plasma gas/l min-1 Ar auxiliary gas/l min-1 Ar carrier gas/l min-1 Sampling depth/mm Measured m/z Dwell time/ms flow rate/ml min-1 Hydride generation parameters— NaBH concentration (%) NaBH flow rate/ml min-1 HNO concentration/mol l-1 HNO LC parameters— Flow rate of mobile phase/ml min-1 Sample injection volume/ml Shim-pack SCR-102H Trifluoroacetic acid (pH 2.1) 1.5 50 or 200 Fig.2 EVect of Na2SO4 concentration on peak shapes of AsV and AsIII. A 0 mg ml-1 Na2SO4; B 40 mg ml-1 Na2SO4; C 400 mg ml-1 Na2SO4. Fig. 3 EVects of NaBH4 concentration on (a) signal peak heights and (b) background equivalent concentrations (BECs). because the transportation of gaseous hydrides from the gas– liquid separator was fairly rapid and the unreacted arsenic species in solution were rapidly removed from the gas–liquid separator by the peristaltic pump. Retention behavior 1 The column used was a sulfonated polystyrene type.The mobile phase was dilute trifluoroacetic acid adjusted to pH 2.1. Since the pK values are reported to be 2.25,27 2.6,28 6.3,28 and 9.2327 for AsV MMA DMA and AsIII respectively and the protonation of DMA is reported to occur at pH 3.85,29 AsV partly exists as an anionic form while MMA and AsIII predominantly exist as neutral forms and DMA exists as a cationic form respectively at pH 2.1. A chromatogram of AsV AsIII and MMA is shown in Fig. 4. The separation between AsV and AsIII appears to be based on an ion exclusion mechanism. However the separation between AsIII and MMA is thought to be based on hydrophobic adsorption by the polymeric resin. DMA is thought to be retained on the column by electrostatic attraction and hydrophobic adsorption.This retention behavior of AsV AsIII and MMA is unique compared with that reported thus far in which the elution of AsIII is usually earlier than AsV. Attempts to elute MMA and DMA more rapidly by using a higher pH eluent and by adding methanol to the eluent were not successful. Calibration graphs detection limits and repeatability Six-point calibration graphs for AsV AsIII and MMA were obtained by plotting the peak areas against the concentration of arsenic for each species in the range 0–10 ng ml-1. The calibration graphs were linear within this range. The slopes and regression coeYcients of the calibration graphs are given in Table 2. The detection limits defined as three times the standard deviation of the peak areas for seven replicates of the blank are also listed in Table 2.These values are more than 20 times lower than those reported to date.3,4 The use of a high-purity acid as the mobile phase is thought to contribute to this improvement in the detection limits. The repeatability for the three arsenic species was evaluated from five replicates using a standard containing 1 ng ml-1 As of each species. The relative standard deviations are given in Table 2. Recovery test In order to verify the reliability of the present method recovery tests were carried out with diVerent types of water such as riverine water sea-water and tap water by adding standards. JAC0031 and SLRS-1 are riverine reference water samples of the Japanese Society for Analytical Chemistry (JSAC) and of The National Research Council of Canada (NRCC) respectively.CASS-3 is a reference sea-water sample of the NRCC. Each arsenic species standard was added to the reference water samples to increase the concentrations by 1 ng ml-1 As per species. The averages and standard deviations of four recovery tests are given in Table 3. Since the tap water is chlorinated AsIII added to it is rapidly oxidized to AsV by hypochlorite. Therefore the tap water used for the recovery test was boiled for 10 min to remove chlorine species. Recoveries of the three arsenic species which ranged from 94 to 108% are thought to be acceptable. Fig. 4 Chromatograms for AsV AsIII and MMA. A Blank; B 0.1 ng ml-1 As; C 0.5 ng ml-1 As; D 1.0 ng ml-1 As of each species.653 J. Anal. At. Spectrom. 1999 14 651–655 Table 2 Calibration graphs detection limits and repeatability AsV 349 0.9993 1.1 0.8 Slope/counts per pg ml-1 As Detection limit/pg ml-1 As Regression coeYcient (R2) Repeatability (n=5) at 1 ng ml-1 As (%) Table 3 Recovery of added AsV AsIII and MMA from water samples (%) (n=4 at 1 ng ml-1 As)a MMA AsIII AsV 94.3±1.9 96.7±1.5 101.0±0.5 JAC0031 riverine water 106.3±7.7 SLRS-1 riverine water 105.0±4.0 CASS-3 sea-water 108.0±5.3 Tap water 97.5±0.7 95.7±4.6 95.0±0.5 96.3±2.3 99.5±3.5 107.5±6.3 a±values are standard deviations. Fig. 5 Schematic diagram of LC-HG-ICP-MS with a column switching system. Column switching method to shorten analysis time Approximately 18 min was required to complete one analysis because of the long retention time of MMA.In order to shorten the analysis time a column switching method was applied. A guard column (SCR-102H 5 cm long) and a sixway motorized valve (Rheodyne Model 9750) were inserted before the separation column (SCR-102H 30 cm) as shown in Fig. 5. Initially the valve was positioned so that AsV and AsIII eluted from the guard column were introduced into the separation column. The valve position was then changed at 1.2 min so that MMA eluted from the guard column could be introduced directly into the hydride generator. Finally after the detection of MMA by ICP-MS the valve was returned to the original position at 2.6 min to complete the separation of AsV and AsIII by the separation column.Peak broadening of AsV and AsIII as a result of stopping the flow in the separation column from 1.2 to 2.6 min was negligible. A better separation was obtained for AsV and AsIII because of the longer total column length (35 cm) compared with the result shown in Fig. 4. These modifications permitted the analysis time to be shortened to 9 min. Chromatograms for SLRS-3 and CASS-3 obtained using an injection volume of 200 ml are shown in Fig. 6 along with the standard containing each species at 1 ng ml-1 As. The AsIII peak was not observed in Table 4 Analytical results for various reference waters and tap water (ng ml-1 As) (n=3) AsV JAC0031 riverine water SLRS-1 riverine water SLRS-3 riverine water SLEW-2 estuarine water CASS-3 sea-water Tap water 0.22±0.014a 0.30±0.014 0.49±0.031 0.80±0.065 1.11±0.07 0.10±0.010 aPrecision expressed as the standard deviation.bBelow detection limit. cUncertainties for the certified values are 95% confidence intervals. 654 J. Anal. At. Spectrom. 1999 14 651–655 MMA AsIII 877 0.9999 0.5 2.8 966 0.9998 0.5 2.5 Fig. 6 Top chromatogram of reference sea-water CASS-3; bottom chromatograms of a standard solution containing 1 ng ml-1 As of each species and riverine reference water SLRS-3. The baseline for the chromatogram of CASS-3 has been shifted so that the peak can be compared readily. The backgrounds of the three chromatograms were nearly identical. 3 3 either of the reference water samples because of acidification with HNO for sample storage.It was confirmed that AsIII was rapidly converted to AsV at low concentrations by adding small portions of concentrated HNO solution at 1 ng ml-1 (final HNO concentration 0.1 mol l-1). 3 to the AsIII standard Therefore it is not recommended that HNO3 is added to the sample solutions when the speciation of arsenic is required. The peak of MMA for SLRS-3 was easily distinguishable and the fluctuation of the background for CASS-3 was much smaller compared with the chromatograms reported previously. 3,4 Although care must be exercised in comparing the chromatograms obtained with the diVerent ICP-MS instruments it appears that the smaller fluctuations and the less noisy background observed here are due to the robustness of the ion exclusion column and the high purity of the mobile phase.It has also been reported that since the high chloride concentration of a reference sea-water sample (NASS-4 of the NRCC salinity 31.3‰) causes peak splitting for MMA the sample must be diluted 1+3 with distilled water in order to be analyzed accurately with an anion exchange column.4 No such problems were encountered with the present method. Certified value MMA AsIII 0.28±0.04c 0.55±0.08 0.72±0.05 0.792±0.082 1.09±0.07 — <d.l. 0.05±0.008 0.08±0.010 0.03±0.005 0.01±0.005 0.06±0.012 <d.l.b <d.l. <d.l. <d.l. <d.l. <d.l. Since DMA could not be eluted from the 5 cm guard column DMA seemed to be retained strongly on the column.Neither a deterioration of the column performance nor elevated background signals due to the retained DMA were observed during the experimental period viz. about 5 months but the guard column should be replaced if such phenomena are observed. Analysis of reference water samples In order to validate the present method three riverine reference water samples (JAC0031 SLRS-1 and SLRS-3) one estuarine reference water sample (SLEW-2 salinity 11.6‰) and one reference sea-water sample (CASS-3 salinity 30.2‰) were analyzed. Analytical results are shown in Table 4 along with the certified values. Arsenite (AsIII) was not detected in any of the samples because of the acidification with HNO3 as mentioned above.The sum of the concentrations of AsV and MMA for CASS-3 and SLEW-2 showed good agreement with the certified values. In contrast those for riverine waters were out of the range of the certified values. In particular the deviations from the certified values were large for SLRS-1 and SLRS-3. The reason for this is not clear at present but one possible reason might be the existence of other species such as DMA which cannot be detected with the present method. A relatively large peak of DMA as large as that of AsV was observed in the SLRS-2 riverine reference water sample (NRCC).3 Conclusions A highly sensitive and robust method for the speciation of arsenic in water was developed by coupling ion exclusion chromatography to ICP-MS with hydride generation as the sample introduction technique.The separation based on ion exclusion and hydrophobic adsorption permitted a unique elution order of AsV>AsIII&MMA by using only a dilute acid as the eluent. The use of a high-purity acid decreased the background level and improved the detection limit. The retention times of the three species were not aVected by the seawater matrix. The method was free from interference from chloride ions and was easily applied to sea-water samples. The main disadvantage of the present method is its failure to elute DMA. This is probably because DMA is retained on the column via hydrophobic adsorption on the polystyrene resin. The use of a more hydrophilic resin should help to decrease this eVect.Methods for the purification of sodium tetrahydroborate should be pursued to achieve a better detection limit since the detection limit was determined by impurities in this reagent. References 1 M. Vahter and E. Marafante in Metal Ions in Biological Systems. Vol. 29. Biological Properties of Metal Alkyl Derivatives ed. H. Sigel and A. Sigel Marcel Dekker New York 1993 p. 161. 2 H. Yamauchi and B. A. Fowler in Arsenic in the Environment Part II ed. J. O. Nriagu Wiley New York 1994 p. 35. 3 C.-J. Hwang and S.-J. Jiang Anal. Chim. Acta 1994 289 205. 4 M. L. Magnuson J. T. Creed and C. A. BrockhoV J. Anal. At. 5 S. J. Hill M. J. Bloxham and P. J. Worsfold J. Anal. At. 6 O. F. X. Donard and E. M. Martin Trends Anal. Chem. 1992 Spectrom.1996 11 893. Spectrom. 1993 8 499. 11 17. 7 H. Tao Bunseki Kagaku 1997 46 239. 8 T. Nakahara Bunseki Kagaku 1997 46 513. 9 M. L. Magnuson J. T. Creed and C. A. BrockhoV J. Anal. At. Spectrom. 1997 12 689. 10 M. A. Lo�pez-Gonza�lves M. M. Go�mez C. Ca�mara and M. A. Palacios J. Anal. At. Spectrom. 1994 9 291. 11 X.-C. Le W. R. Cullen and K. J. Reimer Talanta 1994 41 495. 12 R. H. Atallah and D. A. Kalman Talanta 1991 38 167. 13 A. G. Howard and L. E. Hunt Anal. Chem. 1993 65 2995. 14 S. A. Pergantis E. M. Heithmar and T. A. Hinners Anal. Chem. 1995 67 4530. 15 J. J. Thompson and R. S. Houk Anal. Chem. 1986 58 2541. 16 S. Saverwyns X. Zhang F. Vanhaecke R. Cornelis L. Moens and R. Dams J. Anal. At. Spectrom. 1997 12 1047. 17 S. C. K. Shum R. Neddersen and R. S. Houk Analyst 1992 117 577. 18 D. Beauchemin K. W. M. Siu J. W. McLaren and S. S. Berman J. Anal. At. Spectrom. 1989 4 285. 19 P. Thomas and K. Sniatecki J. Anal. At. Spectrom. 1995 10 615. 20 Y. Shibata and M. Morita Anal. Sci. 1989 5 107. 21 D. Beauchemin M. E. Bednas S. S. Berman J. W. McLaren K. W. M. Siu and R. S. Sturgeon Anal. Chem. 1988 60 2209. 22 S. L. Cleland L. K. Olson J. A. Caruso and J. M. Carey J. Anal. At. Spectrom. 1994 9 975. 23 E. H. Larsen G. Pritzl and S. H. Hansen J. Anal. At. Spectrom. 1993 8 557. 24 E. H. Larsen G. Pritzl and S. H. Hansen J. Anal. At. Spect93 8 1075. 25 S. Branch L. Ebdon and P. O’Neill J. Anal. At. Spectrom. 1994 9 33. 26 H. Tao V. J. Boyko and J. W. McLaren Spectrochim. Acta Part B 1993 48 1339. 27 CRC Handbook of Chemistry and Physics ed. R. C. Weast CRC Press Boca Raton FL 66th edn. 1985 p. D-163. 28 Y. M. Liu M. L. F. Sa�nchez E. B. Gonza�les and A. Sanz-Medel J. Anal. At. Spectrom. 1993 8 815. 29 Y. Shibata M. Morita and J. S. Edmonds Agric. Biol. Chem. 1987 51 391. Paper 8/01132F 655 J. Anal. At. Spectrom. 1999 14 651&ndash